Add Connect To Individual Overmind Processes Via tmux as a tmux TIL

Add Output Query Result In Nicely Formatted Table as a SQLite TIL
Add Register SQLite Adapter To Serialize Datetimes as a Python TIL
2026-09-02 09:41:47 +00:00 · 2026-08-30 08:19:24 -05:00 · 2026-08-29 21:05:44 -05:00 · 2026-08-28 13:46:28 -05:00 · 2026-08-27 14:25:05 -05:00 · 2026-08-25 20:44:10 -05:00
39 changed files with 1949 additions and 1 deletions
@@ -10,7 +10,7 @@ working across different projects via [VisualMode](https://www.visualmode.dev/).

 For a steady stream of TILs, [sign up for my newsletter](https://visualmode.kit.com/newsletter).

-_1839 TILs and counting..._
+_1877 TILs and counting..._

 See some of the other learning resources I work on:

@@ -148,6 +148,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Access A Value Logged To The Console](chrome/access-a-value-logged-to-the-console.md)
 - [Chrome Supports Many Unix Keyboard Shortcuts](chrome/chrome-supports-many-unix-keyboard-shortcuts.md)
 - [Copy Some Data From The Console](chrome/copy-some-data-from-the-console.md)
+- [Duplicate Current Browser Tab](chrome/duplicate-current-browser-tab.md)
 - [Duplicate The Current Tab](chrome/duplicate-the-current-tab.md)
 - [Easier Access To Network Throttling Controls](chrome/easier-access-to-network-throttling-controls.md)
 - [Keybinding To Focus The Address Bar](chrome/keybinding-to-focus-the-address-bar.md)
@@ -171,6 +172,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Monitor Usage Limits From CLI](claude-code/monitor-usage-limits-from-cli.md)
 - [Open Current Prompt In Default Editor](claude-code/open-current-prompt-in-default-editor.md)
 - [Resume Specific Session](claude-code/resume-specific-session.md)
+- [Set Permission Mode When Starting Session](claude-code/set-permission-mode-when-starting-session.md)
 - [Stash The Current Prompt To Send Another First](claude-code/stash-the-current-prompt-to-send-another-first.md)

 ### Clojure
@@ -402,6 +404,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [List All Files Added During Span Of Time](git/list-all-files-added-during-span-of-time.md)
 - [List All Files Changed Between Two Branches](git/list-all-files-changed-between-two-branches.md)
 - [List All Git Aliases From gitconfig](git/list-all-git-aliases-from-gitconfig.md)
+- [List And Count All Posts In TIL Repo](git/list-and-count-all-posts-in-til-repo.md)
 - [List Branches That Contain A Commit](git/list-branches-that-contain-a-commit.md)
 - [List Commits On A Branch](git/list-commits-on-a-branch.md)
 - [List Different Commits Between Two Branches](git/list-different-commits-between-two-branches.md)
@@ -410,6 +413,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [List Most Git Commands](git/list-most-git-commands.md)
 - [List Untracked Files](git/list-untracked-files.md)
 - [List Untracked Files For Scripting](git/list-untracked-files-for-scripting.md)
+- [Mark A Release With An Annotated Tag](git/mark-a-release-with-an-annotated-tag.md)
 - [Move The Latest Commit To A New Branch](git/move-the-latest-commit-to-a-new-branch.md)
 - [Override The Global Git Ignore File](git/override-the-global-git-ignore-file.md)
 - [Pick Specific Changes To Stash](git/pick-specific-changes-to-stash.md)
@@ -486,6 +490,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Capture An Output Value For Use In A Later Step](github-actions/capture-an-output-value-for-use-in-a-later-step.md)
 - [Disable A Workflow With The gh CLI](github-actions/disable-a-workflow-with-the-gh-cli.md)
 - [Reference An Encrypted Secret In An Action](github-actions/reference-an-encrypted-secret-in-an-action.md)
+- [Run Schedule Action To Commit Regular Updates](github-actions/run-scheduled-action-to-commit-regular-updates.md)
 - [Trigger A Workflow Via An API Call](github-actions/trigger-a-workflow-via-an-api-call.md)
 - [Use Labels To Block PR Merge](github-actions/use-labels-to-block-pr-merge.md)

@@ -616,6 +621,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Easy Date Comparison With DayJS](javascript/easy-date-comparison-with-dayjs.md)
 - [Expand Emojis With The Spread Operator](javascript/expand-emojis-with-the-spread-operator.md)
 - [Fill An Input With A Ton Of Text](javascript/fill-an-input-with-a-ton-of-text.md)
+- [Filter By Truthy Values With Boolean Function](javascript/filter-by-truthy-values-with-boolean-function.md)
 - [Find The Version Of An Installed Dependency](javascript/find-the-version-of-an-installed-dependency.md)
 - [Find Where Yarn Is Installing Binaries](javascript/find-where-yarn-is-installing-binaries.md)
 - [for...in Iterates Over Object Properties](javascript/for-in-iterates-over-object-properties.md)
@@ -647,6 +653,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Matching Multiple Values In A Switch Statement](javascript/matching-multiple-values-in-a-switch-statement.md)
 - [Mock A Function With Return Values Using Jest](javascript/mock-a-function-with-return-values-using-jest.md)
 - [New Dates Can Take Out Of Bounds Values](javascript/new-dates-can-take-out-of-bounds-values.md)
+- [`npm run` Has Some Typo Aliases](javascript/npm-run-has-some-typo-aliases.md)
 - [Numbers Are Empty](javascript/numbers-are-empty.md)
 - [Object Initialization With Shorthand Property Names](javascript/object-initialization-with-shorthand-property-names.md)
 - [Obtain Undefined Value With The Void Operator](javascript/obtain-undefined-value-with-the-void-operator.md)
@@ -732,6 +739,7 @@ If you've learned something here, support my efforts writing daily TILs by
 ### LLM

 - [Count Number Of Tokens In A File](llm/count-number-of-tokens-in-a-file.md)
+- [Include A File With Message To `ant`](llm/include-a-file-with-message-to-ant.md)
 - [Send cURL To Claude Text Completion API](llm/send-curl-to-claude-text-completion-api.md)
 - [Use The llm CLI With Claude Models](llm/use-the-llm-cli-with-claude-models.md)

@@ -781,6 +789,7 @@ If you've learned something here, support my efforts writing daily TILs by

 ### Math

+- [Convert Arbitrary Number To Probability With Sigmoid](math/convert-arbitrary-number-to-probability-with-sigmoid.md)
 - [Generate Permutations Of All Valid 9-ball Racks](math/generate-permutations-of-all-valid-9-ball-racks.md)

 ### Mise
@@ -1008,6 +1017,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Restarting Sequences When Truncating Tables](postgres/restarting-sequences-when-truncating-tables.md)
 - [Salt And Hash A Password With pgcrypto](postgres/salt-and-hash-a-password-with-pgcrypto.md)
 - [Send A Command To psql](postgres/send-a-command-to-psql.md)
+- [Sequence Side-Effect When Rolling Back Inserts](postgres/sequence-side-effect-when-rolling-back-inserts.md)
 - [Set Inclusion With hstore](postgres/set-inclusion-with-hstore.md)
 - [Set A Seed For The Random Number Generator](postgres/set-a-seed-for-the-random-number-generator.md)
 - [Set A Statement Timeout Threshold For A Session](postgres/set-a-statement-timeout-threshold-for-a-session.md)
@@ -1071,6 +1081,8 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Access Most Recent Return Value In REPL](python/access-most-recent-return-value-in-repl.md)
 - [Access SQLite Result Values By Name With Row Factory](python/access-sqlite-result-values-by-name-with-row-factory.md)
 - [Access Variables Outside Loop Scope](python/access-variables-outside-loop-scope.md)
+- [Annotate Return Type On Enum Function](python/annotate-return-type-on-enum-function.md)
+- [Another Way To Mark Keyword-Only Dataclass Fields](python/another-way-to-mark-keyword-only-dataclass-fields.md)
 - [Argument Defaults Are Evaluated When Function Is Defined](python/argument-defaults-are-evaluated-when-function-is-defined.md)
 - [Assert Is Only A Development Check](python/assert-is-only-a-development-check.md)
 - [Avoid Modification With Frozen Dataclass](python/avoid-modification-with-frozen-dataclass.md)
@@ -1078,23 +1090,31 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Break Debugger On First Line Of Program](python/break-debugger-on-first-line-of-program.md)
 - [Check If Package Is Installed With Pip](python/check-if-package-is-installed-with-pip.md)
 - [Check Precondition Before Click Arg Parsing](python/check-precondition-before-click-arg-parsing.md)
+- [Combine StrEnum With Pydantic For Union Type](python/combine-strenum-with-pydantic-for-union-type.md)
 - [Commit Writes From Executed SQLite Statements](python/commit-writes-from-executed-sqlite-statements.md)
+- [Configure Other Attributes Of Dataclass Field](python/configure-other-attributes-of-dataclass-field.md)
 - [Control Passing Of Time In Tests](python/control-passing-of-time-in-tests.md)
 - [Create A Dummy DataFrame In Pandas](python/create-a-dummy-dataframe-in-pandas.md)
 - [Create A Range Of Descending Values](python/create-a-range-of-descending-values.md)
+- [Dedent Common Whitespace From Multiline String](python/dedent-common-whitespace-from-multiline-string.md)
 - [Deduplicate A List Into A Tuple](python/deduplicate-a-list-into-a-tuple.md)
 - [Define Sequence Of Tests With Parametrize Decorator](python/define-sequence-of-tests-with-parametrize-decorator.md)
 - [Define Typed Class Interface With Protocol](python/define-typed-class-interface-with-protocol.md)
 - [Dunder Methods](python/dunder-methods.md)
 - [Easy Key-Value Aggregates With defaultdict](python/easy-key-value-aggregates-with-defaultdict.md)
 - [Enable Pyright Type Checking In Cursor](python/enable-pyright-type-checking-in-cursor.md)
+- [Escape Curly Braces Within Formatted String](python/escape-curly-braces-within-formatted-string.md)
+- [Experiment With SQLite Queries In Memory](python/experiment-with-sqlite-queries-in-memory.md)
 - [Force Remaining Arguments To Be Named](python/force-remaining-arguments-to-be-named.md)
+- [Generate Sample PDFs With ReportLab](python/generate-sample-pdfs-with-reportlab.md)
 - [Get Absolute Seconds From `timedelta` Object](python/get-absolute-seconds-from-timedelta-object.md)
 - [Get Quotient And Remainder In One Operation](python/get-quotient-and-remainder-in-one-operation.md)
 - [Globally Install CLI Tool With UV](python/globally-install-cli-tool-with-uv.md)
+- [Initialize A PyTorch Tensor](python/initialize-a-pytorch-tensor.md)
 - [Install With PIP For Specific Interpreter](python/install-with-pip-for-specific-interpreter.md)
 - [Iterate First N Items From Enumerable](python/iterate-first-n-items-from-enumerable.md)
 - [Iterate Over A Dictionary](python/iterate-over-a-dictionary.md)
+- [Join A List Of Strings](python/join-a-list-of-strings.md)
 - [Keep A Tally With collections.Counter](python/keep-a-tally-with-collections-counter.md)
 - [Lint And Format Project With Ruff](python/lint-and-format-project-with-ruff.md)
 - [Load A File Into The Python REPL](python/load-a-file-into-the-python-repl.md)
@@ -1102,14 +1122,19 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Make Dataclass Sortable By Specific Field](python/make-dataclass-sortable-by-specific-field.md)
 - [Make Secure Temp File For Atomic Write](python/make-secure-temp-file-for-atomic-write.md)
 - [Override The Boolean Context Of A Class](python/override-the-boolean-context-of-a-class.md)
+- [Override Timezone That All Tests Run In](python/override-timezone-that-all-tests-run-in.md)
 - [Parse Relative Time To datetime Object](python/parse-relative-time-to-datetime-object.md)
+- [Publish A Package To A Test Env As A Dry Run](python/publish-a-package-to-a-test-env-as-a-dry-run.md)
 - [Reclassify Certain Packages As Dev Dependencies](python/reclassify-certain-packages-as-dev-dependencies.md)
+- [Register SQLite Adapter To Serialize Datetimes](python/register-sqlite-adapter-to-serialize-datetimes.md)
 - [Resurface Exceptions Swallowed By Click Under Test](python/resurface-exceptions-swallowed-by-click-under-test.md)
 - [Select Implementation With Class Registry](python/select-implementation-with-class-registry.md)
 - [Set Up Pyright Type Checking In GitHub](python/set-up-pyright-type-checking-in-github.md)
 - [Skip Specific Pytest Test Cases](python/skip-specific-pytest-test-cases.md)
 - [Sort A List Of Dataclass Instances](python/sort-a-list-of-dataclass-instances.md)
 - [Sort Normalized Version Of Data](python/sort-normalized-version-of-data.md)
+- [Specify Parameter Name For Click Option](python/specify-parameter-name-for-click-option.md)
+- [Start Jupyter Notebook With Extra Packages](python/start-jupyter-notebook-with-extra-packages.md)
 - [Start The Debugger When A Test Errors](python/start-the-debugger-when-a-test-errors.md)
 - [Store And Access Immutable Data In A Tuple](python/store-and-access-immutable-data-in-a-tuple.md)
 - [Strictly Separate Positional And Keyword Arguments](python/strictly-separate-positional-and-keyword-arguments.md)
@@ -1208,6 +1233,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Get The Current Time](rails/get-the-current-time.md)
 - [Grab A Random Record From The Database](rails/grab-a-random-record-from-the-database.md)
 - [Halt ActionMailer Delivery With Callback](rails/halt-action-mailer-delivery-with-callback.md)
+- [Handle Bad Numerical Amounts With BigDecimal](rails/handle-bad-numerical-amounts-with-big-decimal.md)
 - [Handle Named Arguments In A Rake Task](rails/handle-named-arguments-in-a-rake-task.md)
 - [Hash Slicing](rails/hash-slicing.md)
 - [Ignore Poltergeist JavaScript Errors](rails/ignore-poltergeist-javascript-errors.md)
@@ -1251,6 +1277,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Read-Only Models](rails/read-only-models.md)
 - [Rebuild Tailwind Bundle For Dev Server](rails/rebuild-tailwind-bundle-for-dev-server.md)
 - [Remove A Database Column From A Table](rails/remove-a-database-column-from-a-table.md)
+- [Remove Blank Values From A Hash](rails/remove-blank-values-from-a-hash.md)
 - [Remove The Default Value On A Column](rails/remove-the-default-value-on-a-column.md)
 - [Render An Alternative ActionMailer Template](rails/render-an-alternative-action-mailer-template.md)
 - [Render The Response Body In Controller Specs](rails/render-the-response-body-in-controller-specs.md)
@@ -1534,6 +1561,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [OpenStruct Has Bad Performance Characteristics](ruby/open-struct-has-bad-performance-characteristics.md)
 - [Or Operator Precedence](ruby/or-operator-precedence.md)
 - [Output Bytecode For A Ruby Program](ruby/output-bytecode-for-a-ruby-program.md)
+- [Override Output Field Separator](ruby/override-output-field-separator.md)
 - [Override The Initial Sequence Value](ruby/override-the-initial-sequence-value.md)
 - [Parallel Bundle Install](ruby/parallel-bundle-install.md)
 - [Parse JSON Into An OpenStruct](ruby/parse-json-into-an-open-struct.md)
@@ -1624,8 +1652,11 @@ If you've learned something here, support my efforts writing daily TILs by

 ### SQLite

+- [Add Check Constraint To Existing Column](sqlite/add-check-constraint-to-existing-column.md)
 - [Display Results In Readable Column Format](sqlite/display-results-in-readable-column-format.md)
 - [Explore The Database Schema](sqlite/explore-the-database-schema.md)
+- [Manage Lightweight Schema Migrations With `user_version`](sqlite/manage-lightweight-schema-migrations-with-user-version.md)
+- [Output Query Result In Nicely Formatted Table](sqlite/output-query-result-in-nicely-formatted-table.md)

 ### Streaming

@@ -1655,6 +1686,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Change Base Directory Of Existing Session](tmux/change-base-directory-of-existing-session.md)
 - [Change Base Directory Without Detaching](tmux/change-base-directory-without-detaching.md)
 - [Change The Default Prefix Key](tmux/change-the-default-prefix-key.md)
+- [Connect To Individual Overmind Processes Via tmux](tmux/connect-to-individual-overmind-processes-via-tmux.md)
 - [Create A Named tmux Session](tmux/create-a-named-tmux-session.md)
 - [Create A New Session In A New Server](tmux/create-a-new-session-in-a-new-server.md)
 - [Cycle Through Layouts](tmux/cycle-through-layouts.md)
@@ -1727,6 +1759,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Check SSH Key Fingerprints Of Known Hosts](unix/check-ssh-key-fingerprints-of-known-hosts.md)
 - [Check The Current Working Directory](unix/check-the-current-working-directory.md)
 - [Check The Installed OpenSSL Version](unix/check-the-installed-openssl-version.md)
+- [Check What Is Inside A Zip File](unix/check-what-is-inside-a-zip-file.md)
 - [Clear The Screen](unix/clear-the-screen.md)
 - [Combine All My TILs Into A Single File](unix/combine-all-my-tils-into-a-single-file.md)
 - [Command Line Length Limitations](unix/command-line-length-limitations.md)
@@ -1747,6 +1780,7 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Curling For Headers](unix/curling-for-headers.md)
 - [Curling With Basic Auth Credentials](unix/curling-with-basic-auth-credentials.md)
 - [Deduplicate List While Preserving Original Order](unix/deduplicate-list-while-preserving-original-order.md)
+- [Delete Empty Files With Find](unix/delete-empty-files-with-find.md)
 - [Determine ipv4 And ipv6 Public IP Addresses](unix/determine-ipv4-and-ipv6-public-ip-addresses.md)
 - [Diff Two Files In Unified Format](unix/diff-two-files-in-unified-format.md)
 - [Different Ways To Generate A v4 UUID](unix/different-ways-to-generate-a-v4-uuid.md)
@@ -2113,9 +2147,11 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Get Your Public IP Address](workflow/get-your-public-ip-address.md)
 - [Import A Github Project Into CodeSandbox](workflow/import-a-github-project-into-codesandbox.md)
 - [Interactively Kill A Process With fkill](workflow/interactively-kill-a-process-with-fkill.md)
+- [Move YouTube Subtitles Out Of The Way](workflow/move-youtube-subtitles-out-of-the-way.md)
 - [Open Slack's Keyboard Shortcuts Reference Panel](workflow/open-slacks-keyboard-shortcuts-reference-panel.md)
 - [Pop Videos Out As Picture-in-Picture](workflow/pop-videos-out-as-picture-in-picture.md)
 - [Prune The Excess From node_modules](workflow/prune-the-excess-from-node-modules.md)
+- [Remove Pages From A PDF](workflow/remove-pages-from-a-pdf.md)
 - [Rotate An Image To Be Oriented Upright](workflow/rotate-an-image-to-be-oriented-upright.md)
 - [See Overlaps For A Set Of Time Zones](workflow/see-overlaps-for-a-set-of-time-zones.md)
 - [Send A Message To A Discord Channel](workflow/send-a-message-to-a-discord-channel.md)
@@ -2168,7 +2204,9 @@ If you've learned something here, support my efforts writing daily TILs by
 - [Create And Jump Into A Directory](zsh/create-and-jump-into-a-directory.md)
 - [Link A Scalar To An Array](zsh/link-a-scalar-to-an-array.md)
 - [List Available Zle Keybindings](zsh/list-available-zle-keybindings.md)
+- [Open Current Command Line Prompt In Editor](zsh/open-current-command-line-prompt-in-editor.md)
 - [Use A Space To Exclude Command From History](zsh/use-a-space-to-exclude-command-from-history.md)
+- [Where And Which Are Whence](zsh/where-and-which-are-whence.md)

 ## Usage

@@ -0,0 +1,24 @@
+# Duplicate Current Browser Tab
+
+Let's say I'm on a specific page within a web app. Maybe I'm typing out a
+comment on that page. Before I can hit submit, I need to reference something
+elsewhere in the app, perhaps on the previous page.
+
+I can `cmd+t` to open a new tab, type out the URL, hit enter, and then navigate
+around until I find the page I'm looking for. I do this exact thing sometimes,
+but it feels slow and clunky.
+
+The other way I accomplish this which feels way smoother is to _duplicate the
+current tab_. I can do that by `cmd`-clicking the reload button next to the URL
+bar.
+
+This opens another tab at the same URL with the same navigation history. I can
+switch to that tab and hit the back button to go to the previous page, find the
+thing I wanted to reference, and then return to the previous tab to finish what
+I was doing.
+
+To give a more concrete example: this happens all the time with GitHub PR
+comments where I want to copy the URL or number of a specific PR to reference in
+a comment I'm leaving on the current PR. Duplicating the tab, going _back_ to
+the PR index view, finding the other PR, and copying its URL -- to me that is a
+quicker flow popping open a fresh browser tab.
@@ -0,0 +1,29 @@
+# Set Permission Mode When Starting Session
+
+The way I typically use Claude Code day-to-day is with a couple long-running
+sessions for one to two clones of the project. I start a session with `claude`
+and then hit `shift+tab` until I've toggled it to _auto_ mode. I do tightly
+scoped features and `/clear` the context in between each.
+
+I get used to being in _auto_ mode, so whenever I start a new `claude` session I
+forget to first toggle from _manual_ to _auto_ mode.
+
+This is where the
+[`--permission-mode`](https://code.claude.com/docs/en/permission-modes) flag can
+help. I can start a session directly in _auto_ mode like so:
+
+```bash
+❯ claude --permission-mode auto
+```
+
+Or if I know I want to generate a plan first, I can start it in _plan_ mode.
+
+```bash
+❯ claude --permission-mode plan
+```
+
+There is also the `--dangerously-skip-permissions` flag which is equivalent to
+`--permission-mode bypassPermissions`. I tend to stay away from those unless I'm
+working from a sandboxed dev container.
+
+See `claude --help` for more details.
@@ -0,0 +1,39 @@
+# List And Count All Posts In TIL Repo
+
+I want to be able to reliably list and count all posts in [my TIL
+repo](https://github.com/jbranchaud/til). I do this to check that the count in
+the README is accurate and in [the workflow
+script](https://github.com/jbranchaud/jbranchaud/blob/71cba39dffb2bff68bf16d8895e435065e400250/scripts/update_tils.py#L101)
+that powers [my GitHub Profile
+README](https://github.com/jbranchaud/jbranchaud). In the past, I've used
+pattern matching on the listing of all TILs in the
+[README.md](https://github.com/jbranchaud/til/blob/master/README.md). That is
+error prone and has required me to use two different markdown list styles.
+
+A better approach is to ask `git` how many posts it currently has under version
+control. I use a consistent directory structure where each TIL post is a
+markdown file that is nested within a single category directory.
+
+```bash
+❯ git ls-files -- */*.md
+ack/ack-bar.md
+ack/case-insensitive-search.md
+ack/list-available-file-types.md
+...
+```
+
+Using `git ls-files` has the added benefit of only listing files that are
+currently checked in to the project. So if I run this locally, it won't pick up
+a draft post that hasn't been committed yet.
+
+I can then pipe this to `wc -l` (count of all lines) to produce the count:
+
+```bash
+❯ git ls-files -- */*.md | wc -l | xargs
+1850
+```
+
+Note: the empty `xargs` is a trick to trim the whitespace padding that `wc`
+introduces.
+
+See `man git-ls-files` for more details.
@@ -0,0 +1,74 @@
+# Mark A Release With An Annotated Tag
+
+There are two kinds of tags in Git -- lightweight tags and annotated tags.
+
+The [`git-tag` docs](https://git-scm.com/docs/git-tag) explain the distinction:
+
+> Annotated tags are meant for release while lightweight tags are meant for
+> private or temporary object labels.
+
+When an annotated tag is created, a _tag object_ is created which has a creation
+timestamp, a "tagger" (who created it), a message, and potentially a signature
+if GPG commit signing is configured.
+
+I can create an annotated tag for the `HEAD` commit like so:
+
+```bash
+❯ git tag -a v0.1.0 -m "Release v0.1.0"
+```
+
+I can then inspect what was created in a number of ways using `git tag --list`,
+`git show`, `git cat-file`, and `git log --show-signature`.
+
+```bash
+❯ git tag --list
+v0.1.0
+
+❯ git show --no-patch v0.1.0
+tag v0.1.0
+Tagger: jbranchaud <jbranchaud@gmail.com>
+Date:   Sun, 2 Aug 2026 13:57:37 -0500
+
+Release v0.1.0
+
+commit 8a533ecfda526ebd1a4695639830f5620dd8572d (HEAD -> main, tag: v0.1.0, origin/main, origin/HEAD)
+Author: jbranchaud <jbranchaud@gmail.com>
+Date:   Sun, 2 Aug 2026 12:51:41 -0500
+
+    Add changelog with v0.1.0 release changes documented
+
+❯ git cat-file -t v0.1.0
+tag
+
+❯ git cat-file -p v0.1.0
+object 8a533ecfda526ebd1a4695639830f5620dd8572d
+type commit
+tag v0.1.0
+tagger jbranchaud <jbranchaud@gmail.com> 1785697057 -0500
+
+Release v0.1.0
+
+❯ git log --show-signature
+commit 8a533ecfda526ebd1a4695639830f5620dd8572d (HEAD -> main, tag: v0.1.0, origin/main, origin/HEAD)
+gpg: Signature made Sun Aug  2 13:57:09 2026 CDT
+gpg:                using RSA key B2570A9DA3E2A537781501B11A8656918A8D016B
+gpg: Good signature from "jbranchaud <jbranchaud@gmail.com>" [ultimate]
+Author: jbranchaud <jbranchaud@gmail.com>
+Date:   Sun, 2 Aug 2026 12:51:41 -0500
+
+    Add changelog with v0.1.0 release changes documented
+
+...
+```
+
+This tag will be included in a push when I run either of the following:
+
+```bash
+❯ git push origin main
+❯ git push origin v0.1.0
+```
+
+This tag, which is now tied to a _release_, can be seen at [_Releases /
+v0.1.0_](https://github.com/jbranchaud/py-vmt/releases/tag/v0.1.0) on GitHub.
+
+See `man git-tag` for more details.
@@ -0,0 +1,58 @@
+# Run Scheduled Action To Commit Regular Updates
+
+The quintessential example of what GitHub Actions are used for is running CI
+tasks like the test suite, type checker, linter, etc. Each CI step runs to
+completion and either passes or fails and you see the results in the GitHub PR
+interface. That's a great use case, but far from the only one. GitHub Actions
+are a much more general-purpose execution environment that can be used for much
+more.
+
+Here is a recent example that expanded my mind a bit on what is possible with
+GitHub actions. Imagine a GitHub Action that is scheduled to run once a day, it
+executes a script that might make changes to the repo itself (e.g. the
+`README.md`), and then commits those changes (self-updating the repo).
+
+Here is a minimal version of a workflow that does that:
+
+```yaml
+on:
+  schedule:
+    - cron: '17 11 * * *' # daily at 11:17 UTC (06:17 CT)
+  workflow_dispatch:
+
+permissions:
+  contents: write
+
+jobs:
+  update:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4 # get the repo + push credentials
+
+      - run: ./bin/regenerate_readme.sh # apply updates to README
+
+      - run: |
+          git config user.name  "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+          git add -A README.md
+          git diff --staged --quiet || {
+            git commit -m "chore: regenerate README with latest updates"
+            git push
+          } # commit updates to README if there are any, otherwise no-op
+```
+
+Once a day at 11:17 UTC, GitHub Actions will run this workflow which is made up
+of a single job with permissions to write to the repo.
+
+- First, it checks out the repo in the job container which includes push
+credentials.
+- Then it runs a script that might update the README (e.g. maybe based on other
+things that have since been committed to the repo).
+- Last, it attempts to stage any changes the script made to the README. If there
+are any, then it will commit them with that generic commit message. Finally it
+will push that commit to the main branch.
+
+If this all sounds a bit untethered from a real-world example, then take a look
+at how I use this exact pattern to [apply daily updates to my GitHub Profile
+README](https://github.com/jbranchaud/jbranchaud/blob/main/.github/workflows/update-tils.yml)
+based on the latest [TILs](https://github.com/jbranchaud/til) I have written.
@@ -0,0 +1,42 @@
+# Filter By Truthy Values With Boolean Function
+
+The `Boolean` function (not to be confused with the `Boolean` constructor)
+evaluates any given value to its [boolean
+coercion](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Boolean#boolean_coercion).
+
+```javascript
+> Boolean(0)
+false
+> Boolean(1)
+true
+> Boolean(null)
+false
+> Boolean([])
+true
+```
+
+One way that this can be put to use is as a _boolean identity function_ for
+passing to other functions like `filter`.
+
+```javascript
+> [0, 1, "", [], "four", null, "six", undefined, 7].filter(Boolean)
+[ 1, [], 'four', 'six', 7 ]
+```
+
+This filters out all the non-truthy values from a list.
+
+Let's say I'm building a list of nav items that will be rendered to the UI for a
+specific user. Based on permissions or feature flags, certain nav items may not
+be available. Those "empty" entries can be filtered out in this way.
+
+```javascript
+nav_items = [
+  { label: "Home", href: "/" },
+  isSystemAdmin && { label: "System", "/system" },
+  featureEnabled(user, "api") && { label: "API", "/api" },
+].filter(Boolean)
+```
+
+If any of those conditional nav items evaluate to `false`, then they will be
+filtered out. The resulting `nav_items` array is a clean list of actual nav
+items I want to render.
@@ -0,0 +1,62 @@
+# `npm run` Has Some Typo Aliases
+
+The developers of the `npm` CLI know that sometimes we are trying to run
+commands in a hurry. It's easy to be trying to type `npm run` and instead type
+`npm rum` or `npm urn`. No worries though, the command will still work.
+
+If I run `npm help run`, I'll see a manpage that opens with the following:
+
+```
+NPM-RUN(1)                                                                    NPM-RUN(1)
+
+NAME
+       npm-run - Run arbitrary package scripts
+
+   Synopsis
+         npm run <command> [-- <args>]
+
+         aliases: run-script, rum, urn
+
+  ...
+```
+
+Notice it lists a few _aliases_ including `rum` and `urn`.
+
+Here are two examples of me running my test suite with `rum` and then `urn`.
+
+```bash
+❯ npm rum test:run
+
+> test:run
+> vitest run
+
+
+ RUN  v3.0.7 /Users/lastword/dev/jbranchaud/still
+
+ ✓ app/javascript/utils/urlUtils.test.js (6 tests) 2ms
+ ✓ app/javascript/utils/clipboardImage.test.js (20 tests) 3ms
+
+ Test Files  2 passed (2)
+      Tests  26 passed (26)
+   Start at  22:07:53
+   Duration  298ms (transform 17ms, setup 0ms, collect 22ms, tests 5ms, environment 270ms, prepare 63ms)
+
+
+❯ npm urn test:run
+
+> test:run
+> vitest run
+
+
+ RUN  v3.0.7 /Users/lastword/dev/jbranchaud/still
+
+ ✓ app/javascript/utils/urlUtils.test.js (6 tests) 3ms
+ ✓ app/javascript/utils/clipboardImage.test.js (20 tests) 3ms
+
+ Test Files  2 passed (2)
+      Tests  26 passed (26)
+   Start at  22:07:58
+   Duration  305ms (transform 20ms, setup 0ms, collect 25ms, tests 6ms, environment 269ms, prepare 61ms)
+```
+
+See `npm help run` for more details.
@@ -0,0 +1,50 @@
+# Include A File With Message To `ant`
+
+The [Anthropic CLI tool](https://github.com/anthropics/anthropic-cli) (`ant`)
+allows including files, such as images, directly in the message being sent to
+the model. This is done with the `@` symbol followed by a relative path to the
+image on the file system.
+
+I have a file `parameterized-behavioral-tests.png` in my current directory that
+I'd like to include. In the `content` array of the message I include an object
+of type `image` along with some `source` metadata. Within `source` the image
+file is referenced with the `data` field.
+
+With that set, I then include a `text` object with my prompt asking for alt text
+of this image.
+
+```bash
+❯ ant messages create \
+  --model claude-opus-5 \
+  --max-tokens 1024 \
+  --message '{role: user, content: [
+    {type: image, source: {type: base64, media_type: image/png, data: "@./parameterized-behavioral-tests.png"}},
+    {type: text, text: "Produce a concise, but descriptive alt text for this image."}
+  ]}' \
+  --transform 'content.#(type=="text").text' --raw-output
+
+A hand-drawn style diagram comparing two testing approaches, split by a vertical
+line.
+
+On the left, under the heading "Behavioral Tests," is a single column of six
+rounded rectangular test rows: four outlined in green with green check-mark
+icons (passing) and two outlined in red with red X icons (failing), each
+containing black scribble lines representing text.
+
+On the right, under the heading "Parameterized Behavioral Tests," a bracketed
+list of three purple parameter symbols — a circle, a triangle, and a diamond —
+sits at the top, with arrows pointing down to three separate columns of six test
+rows each. Every row in a column is tagged with its corresponding parameter
+shape on the right edge. The circle column shows all six rows passing (green
+with check marks). The triangle column shows three passing and three failing
+(red with X marks). The diamond column shows five passing and one failing. The
+illustration conveys that a single behavioral test, when parameterized, expands
+into multiple variants whose pass/fail outcomes can differ per parameter.
+```
+
+This is for [the image toward the top of this post on parameterized tests in
+pytest](https://www.visualmode.dev/parameterize-a-fixture-instead-of-a-test-case-with-pytest).
+The description is a bit wordier than I would have liked, but it is spot on. It
+is both visually descriptive and conveys what is being conceptually illustrated.
+
+This whole thing was 3800 input tokens, 328 output tokens, and cost ~$0.02.
@@ -0,0 +1,45 @@
+# Convert Arbitrary Number To Probability With Sigmoid
+
+A sigmoid function is a useful function in statistics and machine learning for
+converting a number in the range of positive and negative real numbers into a
+value between 0 and 1. Sigmoid functions can be a bit more diverse than this,
+but this is a good basic definition.
+
+Wikipedia defines another characteristic of sigmoid functions:
+
+> A sigmoid function is any mathematical function whose graph has a
+> characteristic S-shaped or sigmoid curve.
+
+This S-shape is because it is asymptotic at the ends allowing it to cover all
+real numbers in either direction.
+
+A common sigmoid function and the one used by [PyTorch's `Sigmoid`](https://docs.pytorch.org/docs/2.13/generated/torch.nn.Sigmoid.html)
+is this exponential form -- `σ(x) = 1 / (1 + exp(-x))`.
+
+Here is what this looks like plotted on a graph:
+
+![sigmoid function plotted on a graph](https://cdn.visualmode.dev/images/3cadb482-matplot-sigmoid-function-graph.png)
+
+This function can be used any time we want to convert an arbitrary number into a
+probability. Large negative numbers will approach 0. Large positive numbers will
+approach 1. Numbers near 0 will settle somewhere in the middle.
+
+Here are a few examples run through PyTorch's `sigmoid` function:
+
+```python
+print("σ(-99) => ", torch.sigmoid(torch.tensor(-99.0)))
+print("σ(99) => ", torch.sigmoid(torch.tensor(99.0)))
+print("σ(0.123) => ", torch.sigmoid(torch.tensor(0.123)))
+print("σ(-2) => ", torch.sigmoid(torch.tensor(-2.0)))
+print("σ(1) => ", torch.sigmoid(torch.tensor(1.0)))
+```
+
+which prints out:
+
+```
+σ(-99) =>  tensor(0.)
+σ(99) =>  tensor(1.)
+σ(0.123) =>  tensor(0.5307)
+σ(-2) =>  tensor(0.1192)
+σ(1) =>  tensor(0.7311)
+```
@@ -0,0 +1,96 @@
+# Sequence Side-Effect When Rolling Back Inserts
+
+What is so powerful about the transaction guarantees of a system like PostgreSQL
+is that I can make all kinds of changes (big or small) within a transaction,
+then roll it back, and it's like nothing ever happened.
+
+This is useful if I want to dry-run an crucial update statement in the
+production environment. I can gauge how many rows I expect to be impacted, run
+the update in a transaction, inspect the state of the table, and rollback if
+needed. It's like the changes never happened.
+
+Rolled back transactions aren't completely free of side-effects. They can leave
+a trace. In this case, that trace is increments to a sequence like you'd find
+attached to a primary key. Each `insert` of a row into a table with an
+autoincrementing sequence will have an effect on that sequence.
+
+First, I'll create a fresh table and check the state of the sequence.
+
+```sql
+> create table books (id bigserial primary key, title text not null, author text not null);
+CREATE TABLE
+
+> select * from books_id_seq;
+┌────────────┬─────────┬───────────┐
+│ last_value │ log_cnt │ is_called │
+├────────────┼─────────┼───────────┤
+│          1 │       0 │ f         │
+└────────────┴─────────┴───────────┘
+(1 row)
+```
+
+Now, I'll start a transaction, do some inserts, and again check the state of the
+sequence.
+
+```sql
+> begin;
+BEGIN
+
+> insert into books (title, author) values ('Python in a Nutshell', 'Martelli et al'), ('Build an LLM', 'Sebastian Raschka'), ('The Joy of Abstraction', 'Eugenia Cheng');
+INSERT 0 3
+
+> select * from books;
+┌────┬────────────────────────┬───────────────────┐
+│ id │         title          │      author       │
+├────┼────────────────────────┼───────────────────┤
+│  1 │ Python in a Nutshell   │ Martelli et al    │
+│  2 │ Build an LLM           │ Sebastian Raschka │
+│  3 │ The Joy of Abstraction │ Eugenia Cheng     │
+└────┴────────────────────────┴───────────────────┘
+(3 rows)
+
+> select * from books_id_seq;
+┌────────────┬─────────┬───────────┐
+│ last_value │ log_cnt │ is_called │
+├────────────┼─────────┼───────────┤
+│          3 │      30 │ t         │
+└────────────┴─────────┴───────────┘
+(1 row)
+```
+
+I will now rollback the transaction. I expect `books` to be empty again. But
+what happens to the state of `books_id_seq`?
+
+```sql
+> rollback;
+ROLLBACK
+
+> select * from books;
+┌────┬───────┬────────┐
+│ id │ title │ author │
+├────┼───────┼────────┤
+└────┴───────┴────────┘
+(0 rows)
+
+> select * from books_id_seq;
+┌────────────┬─────────┬───────────┐
+│ last_value │ log_cnt │ is_called │
+├────────────┼─────────┼───────────┤
+│          3 │      30 │ t         │
+└────────────┴─────────┴───────────┘
+(1 row)
+```
+
+Despite rolling back, the state of `books_id_seq` remains with a `last_value` of
+`3`. This is because a sequence is used outside of transaction isolation.
+Multiple concurrent transactions may both need sequence values. Instead of them
+blocking on each other, the sequence can always be accessed. The tradeoff is
+that the `last_value` of a sequence is always marching forward and gaps will
+happen when transactions are rolled back.
+
+This is normal and expected behavior. It is something to be aware of if you have
+a high-volume system that is doing a ton of inserting and rolling back (I've
+worked on systems that do this). However, the solution there is more to do with
+picking `bigint` (`bigserial`) for your sequence-backed column. That would
+reasonably eliminate concern for running out of sequence values. Or use UUIDs
+instead.
@@ -0,0 +1,61 @@
+# Annotate Return Type On Enum Function
+
+I defined the following `StrEnum` class to capture the canonical set of valid
+values for a _storage format_ in
+[`py-vmt`](https://github.com/jbranchaud/py-vmt). It includes a `default`
+function that can be called to get what the system considers the default storage
+format.
+
+```python
+class StorageFormat(StrEnum):
+    SQLITE = "sqlite"
+    JSON = "json"
+
+    @staticmethod
+    def default():
+        return StorageFormat.SQLITE
+```
+
+This works, but what is missing is a type annotation for the return value of
+`default`.
+
+I'd like to use `StorageFormat` as the return type annotation. However, the
+annotations are eagerly evaluated at the time the function definition is being
+processed. The `StorageFormat` class is not fully evaluated at that point and so
+isn't available.
+
+Instead, what I can do is quote `StorageFormat` as a string to make it a
+_forward reference_.
+
+```python
+class StorageFormat(StrEnum):
+    SQLITE = "sqlite"
+    JSON = "json"
+
+    @staticmethod
+    def default() -> "StorageFormat":
+        return StorageFormat.SQLITE
+```
+
+The string type annotation will be lazily evaluated at a later time when
+`StorageFormat` fully exists. This is behavior that is slated to be deprecated.
+And it is no longer needed as of [Python 3.14 which lazily evaluates all
+annotations by default
+now](https://docs.python.org/3/reference/compound_stmts.html#annotations).
+
+If I want to be more specific with the above annotation, I can import [`Literal`
+from `typing`](https://docs.python.org/3/library/typing.html#typing.Literal) and
+then wrap a specific enum value.
+
+```python
+class StorageFormat(StrEnum):
+    SQLITE = "sqlite"
+    JSON = "json"
+
+    @staticmethod
+    def default() -> "Literal[StorageFormat.SQLITE]":
+        return StorageFormat.SQLITE
+```
+
+Now it will specifically enforce `SQLITE` rather than accepting `SQLITE` or
+`JSON`.
@@ -0,0 +1,47 @@
+# Another Way To Mark Keyword-Only Dataclass Fields
+
+In [Configure Other Attributes Of Dataclass
+Field](configure-other-attributes-of-dataclass-field), I showed how the
+[`dataclasses.field`](https://docs.python.org/3/library/dataclasses.html#dataclasses.field)
+constructor function can be used. One of the parameters I demonstrated was
+`kw_only`. Each field constructed with `kw_only=True` will be required to be
+passed as a keyword-only parameter when constructing an instance of that
+`dataclass`.
+
+Another way to specify keyword-only parameters with `dataclass` fields is to
+segment them with `KW_ONLY`. This sentinel value can be included as a
+pseudo-field where all fields that come after it are treated as keyword-only.
+
+Translating the example from that other post would look like this:
+
+```python
+from dataclasses import dataclass, field, KW_ONLY
+from datetime import datetime
+
+@dataclass
+class Session:
+    start_time: datetime
+    project_name: str
+    _: KW_ONLY
+    tags: list[str] = field(default_factory=list, kw_only=True)
+    end_time: datetime | None = None
+
+    # ...
+
+sesh1 = Session(start1, "my-project", tags=["pytorch", "numpy"])
+sesh2 = Session(start2, "other-project", end_time=datetime.now())
+```
+
+The field whose value is `KW_ONLY` is only used to signal that keyword-only
+boundary. It does not itself become a field of the dataclass.
+
+On the one hand I like this approach because it feels closer to [the way this is
+signaled in standard function definition
+syntax](force-remaining-arguments-to-be-named.md).
+
+```python
+def build_session(start_time, project_name, *, tags, end_time=None)
+```
+
+On the other hand, it feels like magic `dataclass` syntax whereas the
+`kw_only=True` parameter is more explicit.
@@ -0,0 +1,54 @@
+# Combine StrEnum With Pydantic For Union Type
+
+I want a [Pydantic](https://pydantic.dev/docs/validation/latest/get-started/)
+model that I can use to verify the shape and values of user-provided config
+file. Let's say, for now, the only config value supported by the file is
+`storage_format` and it can be one of three values (`json`, `sqlite`,
+`inmemory`).
+
+I can start with a
+[`StrEnum`](https://docs.python.org/3/library/enum.html#enum.StrEnum) class to
+represent the set of valid values for the storage format field.
+
+```python
+from enum import StrEnum
+from typing import Literal
+
+class StorageFormat(StrEnum):
+    SQLITE = "sqlite"
+    JSON = "json"
+    INMEMORY = "inmemory"
+
+    @classmethod
+    def default(cls) -> Literal[StorageFormat.SQLITE]:
+        return cls.SQLITE
+```
+
+Notice I've also provided a class method called `default` for serving up a
+fallback default value.
+
+Here is the corresponding Pydantic `CliConfig` model which uses the
+`StorageFormat` string enum class as the type signature for its `storage_format`
+field.
+
+```python
+from pydantic import BaseModel
+
+class CliConfig(BaseModel):
+    storage_format: StorageFormat = StorageFormat.default()
+```
+
+For `storage_format` to be valid in the context of `CliConfig`, it must be one
+of the values defined by that `StrEnum`. I get auto-complete, I get type
+checking, and I can validate a config file I've just read into memory (in this
+case, using
+[`model_validate_json`](https://pydantic.dev/docs/validation/dev/api/pydantic/base_model/#pydantic.BaseModel.model_validate_json)).
+
+```python
+def _read_config(self) -> CliConfig:
+    if self.exists():
+        raw_json = self.config_file.read_text()
+        return CliConfig.model_validate_json(raw_json)
+
+    return CliConfig()
+```
@@ -0,0 +1,63 @@
+# Configure Other Attributes Of Dataclass Field
+
+I have a basic [`dataclass`](https://docs.python.org/3/library/dataclasses.html)
+defined with a few fields that looks like this:
+
+```python
+from dataclasses import dataclass
+from datetime import datetime
+
+@dataclass
+class Session:
+    start_time: datetime
+    project_name: str
+    tags: list[str]
+    end_time: datetime | None = None
+
+    # ...
+
+new_session = Session(start_time, "my-project", [])
+```
+
+This syntax lets me define the fields that make up arguments to the underlying
+`__init__` function. This only gets me so far though. What if I want `tags` to
+have a default value of `[]`? What if I want to force those optional fields
+(`tags` and `end_time`) to be keyword-only arguments?
+
+To take the field definitions of my `dataclass` a bit further, I need the
+[`field` function](https://docs.python.org/3/library/dataclasses.html#dataclasses.field)
+which lets me configure more aspects of each field.
+
+```python
+from dataclasses import dataclass, field
+from datetime import datetime
+
+@dataclass
+class Session:
+    start_time: datetime
+    project_name: str
+    tags: list[str] = field(default_factory=list, kw_only=True)
+    end_time: datetime | None = field(default=None, kw_only=True)
+
+    # ...
+
+sesh1 = Session(start1, "my-project", tags=["pytorch", "numpy"])
+sesh2 = Session(start2, "other-project", end_time=datetime.now())
+```
+
+If I were to say the default value of `tags` is simply a `[]`, then the
+`dataclass` definition gets evaluated with a reference to that specific instance
+of an empty list. Every new `Session` I create would share and mutate that same
+list. For this reason, `dataclass` detects and warns you away from that.
+
+Instead, I specify `list` as the `default_factory` for `tags`. Each new
+`Session` where `tags` is not provided will be factory'd a new empty list.
+
+For scalar values, like `None`, I can use the `default` parameter instead like
+I've done for `end_time`.
+
+As for marking certain fields as keyword-only, the `kw_only` parameter can be
+overridden with `True`.
+
+See [the docs for `field`](https://docs.python.org/3/library/dataclasses.html#dataclasses.field)
+for all the other configurable options.
@@ -0,0 +1,46 @@
+# Dedent Common Whitespace From Multiline String
+
+I'm adding tests to cover some custom logic in the _usage_ messages that are
+produced by Click [in a CLI I'm building](https://github.com/jbranchaud/py-vmt).
+The usage output is a mutliline string formatted with some indentation. On top
+of that, I'm putting it in a Python test file with some additional syntactical
+indentation.
+
+If this was Ruby, I would reach for the `<<~` heredoc syntax. The closest
+equivalent to that in Python is
+[`textwrap.dedent`](https://docs.python.org/3/library/textwrap.html#textwrap.dedent).
+
+> Remove any common leading whitespace from every line in text.
+>
+> This can be used to make triple-quoted strings line up with the left edge of
+> the display, while still presenting them in the source code in indented form.
+
+```python
+import textwrap
+
+def test_base_help_message():
+    runner = BetterCliRunner()
+
+    help_result = runner.invoke(cli, ["--help"])
+    output = textwrap.dedent("""\
+        Usage: vmt [OPTIONS] COMMAND [ARGS]...
+
+        Options:
+          -v, --verbose  See extra output when running commands
+          --help         Show this message and exit.
+
+        Commands:
+          cancel
+          log
+          start
+          status
+          stop
+    """)
+    assert output in help_result.output
+```
+
+Before adding in `textwrap.dedent`, the test was failing because of bits of
+extra whitespace that wasn't matching up with the actual CLI output.
+
+Note: the leading `\` after the opening `"""` is needed to remove the leading
+newline that would otherwise be included.
@@ -0,0 +1,47 @@
+# Escape Curly Braces Within Formatted String
+
+Python's
+[_f-string_](https://docs.python.org/3/reference/lexical_analysis.html#f-strings)
+allows you to interpolate values into a string literal like so:
+
+```python
+>>> amount = 25
+>>> print(f"Price: ${amount}")
+Price: $25
+```
+
+The `{` and `}` are betting special treatment when `f` precedes the string
+literal. So, what if the resulting string is supposed to have literal `{` and
+`}` characters in it?
+
+```python
+>>> f"JSON opens and closes with curlies: {...}"
+'JSON opens and closes with curlies: Ellipsis'
+```
+
+I need a way of escaping them. The trick is to double them up.
+
+```python
+>>> f"JSON opens and closes with curlies: {{...}}"
+'JSON opens and closes with curlies: {...}'
+```
+
+A more practical example of where this is useful is in formatting an expected
+string of JSON in a test:
+
+```python
+info_result = runner.invoke(cli, ["info", "--json"])
+output = textwrap.dedent(f"""
+{{
+    "config_file": "{config_file}",
+    "data_dir": "{data_dir}",
+    "version": "{VMT_VERSION}",
+    "schema_version": 1
+}}
+""").strip()
+assert output in info_result.output
+```
+
+I'm doing some interpolation at various points in the string literal, but the
+outer curly braces are meant to be literal, the opening and closing of the JSON
+object.
@@ -0,0 +1,44 @@
+# Experiment With SQLite Queries In Memory
+
+Instead of connecting
+[`sqlite3`](https://docs.python.org/3/library/sqlite3.html) to a real file
+(database) like so:
+
+```python
+db_file: Path = data_dir / "sessions.db"
+conn: Connection = sqlite3.connect(db_file)
+```
+
+I can point it to `":memory:"`:
+
+```python
+conn: Connection = sqlite3.connect(":memory:")
+```
+
+One way this can be useful is when experimenting with some DDL (schema-modifying
+SQL statements), e.g. creating a table, adding a column, and so forth.
+
+```python
+>>> import sqlite3
+>>> conn = sqlite3.connect(":memory:")
+>>> conn.execute("""
+... create table projects (
+...     id integer primary key,
+...     name text not null unique,
+...     created_at text not null default (datetime('now')),
+...     updated_at text not null default (datetime('now'))
+... );
+... """)
+<sqlite3.Cursor object at 0x105d6e2c0>
+>>> conn.execute("insert into projects (name) values ('TIL'), ('py-vmt'), ('Pool League Pro');")
+<sqlite3.Cursor object at 0x105d6e340>
+>>> result = conn.execute("select * from projects;")
+>>> rows = result.fetchall()
+>>> rows
+[(1, 'TIL', '2026-08-01 23:59:56', '2026-08-01 23:59:56'), (2, 'py-vmt', '2026-08-01 23:59:56', '2026-08-01 23:59:56'), (3, 'Pool League Pro', '2026-08-01 23:59:56', '2026-08-01 23:59:56')]
+```
+
+This does not create any sort of on-disk version of the database. Two separate
+connections created this way will be independent in-memory database instances.
+This is technically more of [a SQLite feature](https://sqlite.org/inmemorydb.html) than a Python one, but it was in a
+Python context that I found this most useful.
@@ -0,0 +1,44 @@
+# Generate Sample PDFs With ReportLab
+
+The `reportlab` package has a `pdfgen` module that provides a `Canvas` API for
+programmatically constructing a PDF.
+
+I recently needed the most basic use of this to quickly construct a sample PDF
+that I could use as I experimented with [different `qpdf`
+commands](/workflow/remove-pages-from-a-pdf.md). I wanted to generate a 10-page
+PDF where each page included a large font number in the center. This way I could
+clearly see what was the first, second, third, and so on pages. Then I could
+quickly verify that the `qpdf` manipulations I was doing were working as
+expected.
+
+I was able to do this without any setup or install. Using a `uv` one-liner (does
+a 10-line heredoc count as a one-liner?), I generated a PDF with some basic
+canvas rendering on each page.
+
+```python
+❯ uv run --with reportlab python - <<'EOF'
+from reportlab.pdfgen import canvas
+N, W, H = 10, 612, 792
+c = canvas.Canvas("test.pdf", pagesize=(W, H))
+for i in range(1, N + 1):
+    c.setFont("Helvetica-Bold", 220)
+    c.drawCentredString(W / 2, H / 2 - 80, str(i))
+    c.setFont("Helvetica", 28)
+    c.drawCentredString(W / 2, 100, f"page {i} of {N}")
+    c.showPage()
+c.save()
+EOF
+```
+
+The `--with` makes the `reportlab` package available to the command being run.
+The command being run is `python` which evaluates that block of heredoc text.
+
+Within the heredoc, I import `canvas` from the `reportlab` package that has been
+made available. I then create an in-memory PDF of a standard size. Then for-loop
+10 times to build each page of the PDF with a big number in the middle and a
+smaller `page x of N` lower down on the page. Finally, I save the PDF which
+writes it to the directory I executed this whole command from.
+
+I had Claude help with the `canvas` API, however there are some decent
+[`reportlab.pdfgen` docs
+here](https://docs.reportlab.com/reportlab/userguide/ch2_graphics/).
@@ -0,0 +1,81 @@
+# Initialize A PyTorch Tensor
+
+A _tensor_ is an n-dimensional array that takes on a specified shape and holds
+scalars, typically floats, in each of its positions.
+[PyTorch](https://pytorch.org/) has a comprehensive suite of tools for working
+with tensors. To work with tenors, the first thing I need to do is initialize
+one. Here are a handful of ways to create a 2-dimensional tensor (a matrix)
+depending on various needs.
+
+Here is a 3x4 matrix full of zeros:
+
+```python
+>>> torch.zeros(torch.Size((3, 4)))
+tensor([[0., 0., 0., 0.],
+        [0., 0., 0., 0.],
+        [0., 0., 0., 0.]])
+```
+
+Here is a 3x4 matrix full of ones:
+
+```python
+>>> torch.ones(3,4)
+tensor([[1., 1., 1., 1.],
+        [1., 1., 1., 1.],
+        [1., 1., 1., 1.]])
+```
+
+And here is a 3x4 matrix full of a specific other value:
+
+```python
+>>> torch.full((3,4), 13.0)
+tensor([[13., 13., 13., 13.],
+        [13., 13., 13., 13.],
+        [13., 13., 13., 13.]])
+```
+
+PyTorch is very flexible. I can specify the shape of the `tensor` with
+positional arguments or a tuple for the dimensions. I can even construct a
+`torch.Size` object.
+
+Here is an arguably more useful example where the matrix is seeded with random
+values in the range `[0,1)` using
+[`torch.rand`](https://docs.pytorch.org/docs/2.13/generated/torch.rand.html):
+
+```python
+>>> torch.rand(torch.Size((3,4)))
+tensor([[0.4148, 0.8045, 0.3093, 0.3363],
+        [0.0120, 0.7161, 0.1108, 0.5510],
+        [0.4805, 0.9430, 0.2852, 0.0966]])
+```
+
+These could be used as starting weights in a training process that then get
+tweaked over time.
+
+There are other random tensor functions like
+[`torch.randint`](https://docs.pytorch.org/docs/2.13/generated/torch.randint.html)
+and
+[`torch.randn`](https://docs.pytorch.org/docs/2.13/generated/torch.randn.html).
+
+How about a [random permutation](https://docs.pytorch.org/docs/2.13/generated/torch.randperm.html)
+of integers `[0,12)` reshaped into a 3x4 matrix:
+
+```python
+>>> torch.randperm(12).reshape(3,4)
+tensor([[ 0,  3,  1, 11],
+        [ 7,  8,  9,  4],
+        [10,  5,  6,  2]])
+```
+
+And though there are many other ways to initialize a tensor, the last one I will
+show is
+[`torch.eye`](https://docs.pytorch.org/docs/2.13/generated/torch.eye.html) which
+creates a matrix with ones down the diagonal.
+
+```python
+>>> torch.eye(4)
+tensor([[1., 0., 0., 0.],
+        [0., 1., 0., 0.],
+        [0., 0., 1., 0.],
+        [0., 0., 0., 1.]])
+```
@@ -0,0 +1,30 @@
+# Join A List Of Strings
+
+Though joining a list of strings in Python is a basic task, I wanted to write
+about it because it is backward from how it is done in Ruby (which trips me up
+every single time).
+
+So, in Ruby I would do the following:
+
+```ruby
+> character = ["Gimli", "Dwarf", "Fighter", "Lvl 23"]
+=> ["Gimli", "Dwarf", "Fighter", "Lvl 23"]
+> character.join(" ~ ")
+=> "Gimli ~ Dwarf ~ Fighter ~ Lvl 23"
+```
+
+Notice that I call
+[`join`](https://docs.ruby-lang.org/en/master/Array.html#method-i-join) on the
+list of strings, passing it the specific separator that I want to use.
+
+Python does it the other way around:
+
+```python
+>>> character = ["Gimli", "Dwarf", "Fighter", "Lvl 23"]
+>>> " ~ ".join(character)
+'Gimli ~ Dwarf ~ Fighter ~ Lvl 23'
+```
+
+The separator is the object that I call
+[`join`](https://docs.python.org/3/library/stdtypes.html#str.join) on, passing
+it the list of strings that I want to join.
@@ -0,0 +1,61 @@
+# Override Timezone That All Tests Run In
+
+I was building out the initial functionality of a Python project in my local dev
+env. I was writing tests along the way to guard against regressions. Many of the
+tests assert things about time, durations, and formatted timestamps. All was
+well.
+
+Then I was getting the project setup on GitHub with a CI workflow to run the
+test suite. And a ton of tests were failing. It was all the tests dealing with
+time. They were all off by 5 hours.
+
+I recognized that time difference immediately. That's the difference between my
+time here in Chicago and UTC. When I run the tests locally, they are all
+formatted the timezone setting on my machine -- `America/Chicago`. Whereas when
+I run them in this CI environment, they were suddenly being formatted to `UTC`.
+
+What I needed to do was explicitly set the timezone that all the tests would
+always run in, regardless of environment.
+
+I added the following to my `conftest.py` file:
+
+```python
+import os
+import time
+
+
+# Set the timezone that all tests run in
+os.environ["TZ"] = "America/Chicago"
+time.tzset()
+```
+
+First, I set the timezone (`TZ`) in `os.environ` to `America/Chicago`. That's a
+copy of the executing environment that Python makes when it starts up.
+
+Second, because `TZ` may be tied to some cached info under the hood, I have to
+also run `time.tzset()`.
+
+> Initialize, or reinitialize, the local timezone to the value stored in
+> os.environ[‘TZ’]. The TZ environment variable should be specified in standard
+> Unix timezone format as documented in the tzset man page (eg. ‘US/Eastern’,
+> ‘Europe/Amsterdam’).
+
+Now, all of my tests run reliably in `America/Chicago`.
+
+To demonstrate all of this from the Python repl, here is an example:
+
+```python
+>>> from datetime import datetime
+>>> import os
+>>> import time
+>>> os.environ["TZ"] = "America/Los_Angeles"
+>>> time.tzset()
+>>> datetime.now()
+datetime.datetime(2026, 8, 15, 14, 3, 19, 373682)
+>>> os.environ["TZ"] = "America/Chicago"
+>>> time.tzset()
+>>> datetime.now()
+datetime.datetime(2026, 8, 15, 16, 3, 33, 322803)
+```
+
+Notice the two hour difference between those two.
@@ -0,0 +1,35 @@
+# Publish A Package To A Test Env As A Dry Run
+
+As I was preparing to register a new Python package with PyPI and release my
+first version cut, I felt like there were a lot of unknowns. Is my package's
+name going to be valid? How will the package appear in PyPI? Did I configure
+everything correctly? Etc.
+
+It turns out that PyPI has a great way of answering a lot of these questions.
+There is a [`test.pypi.org`](https://test.pypi.org/) site that parrots the
+publishing flow of `pypi.org`. This makes for a great target to do a dry-run
+publishing of a package.
+
+First, I had to go through the same registration flow and 2FA setup as when I
+registered with `pypi.org`.
+
+Second, I deviated from my tag and CI-triggered publishing flow by instead doing
+a one-off run of the `uv publish` command. That requires an API token which I
+generated in the web UI for `test.pypi.org`. I added that to my env as
+`TEST_PYPI_TOKEN`.
+
+I then ran the following command:
+
+```bash
+❯ uv publish --publish-url https://test.pypi.org/legacy/ --token "$TEST_PYPI_TOKEN"
+```
+
+When this first ran for [`py-vmt`](https://github.com/jbranchaud/py-vmt), I got
+an error back from the publishing API telling me the package name is too similar
+to an existing package. I then had to make a few updates across the project to
+rename the published package name to `visualmode-tracker`. Running the `uv
+publish` command again worked with the updated name.
+
+I was then able to go into the web UI and verify everything looked as expected.
+
+I now have the confidence to publish this thing for real to `pypi.org`.
@@ -0,0 +1,93 @@
+# Register SQLite Adapter To Serialize Datetimes
+
+SQLite doesn't have a datetime or timestamp data type, so that information has
+to be stored as `text` or a unix epoch `int`. That means when using Python's
+`sqlite3` module to perform writes I need to tell it what shape to write a
+datetime value.
+
+I could manually convert `datetime` values everywhere they are involved in
+_write_ operations.
+
+```python
+# Prepare `sessions` insert payload
+session_data = {
+    "active": 1 if active else 0,
+    "project_id": project_id,
+    "start_time": datetime.isoformat(session.start_time),
+    "end_time": None,
+}
+
+if session.end_time:
+    session_data["end_time"] = datetime.isoformat(session.end_time)
+
+# Insert the new active session
+cursor = self.conn.execute(
+    """
+    insert into sessions (active, project_id, start_time, end_time)
+    values (:active, :project_id, :start_time, :end_time)
+    returning id;
+""",
+    session_data,
+)
+```
+
+I've used `datetime.isoformat` above which formats `datetime` objects like so:
+
+```python
+>>> datetime.now().isoformat()
+'2026-08-28T11:52:04.709907'
+```
+
+I'd like to make two improvements.
+
+1. I want these `datetime` values to be formatted instead like `2026-08-28T18:15:27.213Z`.
+2. I want `datetime` values to be serialized automatically in the specific shape
+   without having to manually convert them everywhere.
+
+I can achieve both of these things by [registering an adapter with `sqlite3`](https://docs.python.org/3/library/sqlite3.html#how-to-register-adapter-callables)
+that handles the serialize of `datetime` objects.
+
+First, I define a function that can perform the `datetime` to `str` conversion.
+I decided to put this in `db.py` where I have some other database-specific
+functions.
+
+```python
+from datetime import datetime, timezone
+
+def to_db(dt: datetime) -> str:
+    if dt.tzinfo is None or dt.utcoffset() is None:
+        raise ValueError(f"Unable to store naive datetime: {dt!r}")
+    dt = dt.astimezone(timezone.utc)
+    return f"{dt:%Y-%m-%dT%H:%M:%S}.{dt.microsecond // 1000:03d}Z"
+```
+
+Then I register the adapter before creating the connection that gets used for
+database interactions.
+
+```python
+import sqlite3
+from datetime import datetime
+from pathlib import Path
+from sqlite3 import Connection
+
+def initialize_conn(db_file: Path) -> Connection:
+    # register adapters
+    sqlite3.register_adapter(datetime, to_db)
+
+    conn: Connection = sqlite3.connect(db_file)
+    conn.row_factory = sqlite3.Row
+
+    return conn
+```
+
+Then I can run write operations with `datetime` objects knowing they will be
+correctly serialized.
+
+```python
+with self.conn:
+    query = "update sessions set active = :active, end_time = :end_time where active = 1;"
+    self.conn.execute(
+        query,
+        {"active": 0, "end_time": session.end_time},
+    )
+```
@@ -0,0 +1,55 @@
+# Specify Parameter Name For Click Option
+
+[Click](https://click.palletsprojects.com/en/stable/)'s option decorator
+provides a versatile way to define flags for a command. It has good defaults
+that minimize the aspects of a flag that I need to be explicit about.
+
+For example, a boolean `--init` flag for the `config` command could be specified
+like so:
+
+```python
+@cli.command()
+@click.option(
+    "--init",
+    help="Initialize a config file with minimal defaults",
+    is_flag=True,
+)
+@pass_cli
+def config(cli_ctx: CliContext, init: bool):
+    # ...
+```
+
+Notice, in particular, that the flag string (`--init`) that I pass as the first
+argument to `@click.option` has to correspond to the name of the parameter
+`init`. Click passes all the defined options as keyword arguments when invoking
+`config`. If the `init` parameter was changed to `initial`, there would be a
+runtime error like this: `TypeError: config() got an unexpected keyword argument 'init'`.
+
+Like I said though, Click is flexible when I need it to be. I can leave the flag
+name as it is, but specify a different name to be used for the function
+parameter. I found this useful when I realized that as I added support for a
+`--json` flag I was inadvertently superseding the `json` import.
+
+The second positional argument to `@click.option` can be included to rename that
+parameter:
+
+```python
+import json
+
+
+@cli.command()
+@click.option(
+    "--json",
+    "use_json",
+    help="Output all info details in JSON format",
+    is_flag=True,
+)
+@pass_cli
+def info(cli_ctx: CliContext, use_json: bool):
+    # ...
+
+    if use_json:
+        click.echo(json.dumps(info_details, indent=2))
+```
+
+Both of these code blocks are excerpts from my [`py-vmt` time tracker project](https://github.com/jbranchaud/py-vmt).
@@ -0,0 +1,32 @@
+# Start Jupyter Notebook With Extra Packages
+
+I can start up a one-off [Jupyter notebook](https://jupyter.org/) with
+[`uv`](https://docs.astral.sh/uv/) with the following command:
+
+```bash
+❯ uv run --with jupyter jupyter lab 
+```
+
+The `--with` flag is a `uv run` feature for [requesting additional
+dependencies](https://docs.astral.sh/uv/concepts/projects/run/#requesting-additional-dependencies).
+In this case, it includes the `jupyter` package as a dependency which is
+necessary for the command to then run `jupyter lab`.
+
+As I started following along with [Andrej Karpathy's "Neural Networks: Zero to
+Hero" series](https://karpathy.ai/zero-to-hero.html), I quickly ran into an
+issue with missing other dependencies. In the opening video, Andrej is importing
+`numpy` and `matplotlib`. To ensure those packages are also available for import
+in this notebook, I need to `--with` them as well.
+
+```bash
+❯ uv run --with jupyter --with numpy --with matplotlib jupyter lab 
+```
+
+Now, when I execute the following step, I don't have any import issues:
+
+```python
+import math
+import numpy as np
+import matplotlib.pyplot as plt
+%matplotlib inline
+```
@@ -0,0 +1,38 @@
+# Handle Bad Numerical Amounts With BigDecimal
+
+I'm working on a payment page with a backing Rails controller. The user can
+select between their full balance or some partial payment amount. Because this
+form accepts an arbitrary value for the amount, I need to do some server-side
+validation.
+
+While I could parse the `amount` value and handle the exception that gets raised
+on bad numerical values, I'd rather have bad values coerce to `nil` and let
+downstream validations handle it from there.
+
+[`BigDecimal`](https://docs.ruby-lang.org/en/master/BigDecimal.html) can help
+here with the support of its `exception` option.
+
+```ruby
+> BigDecimal('123', exception: false)
+=> 0.123e3
+> BigDecimal('taco', exception: false)
+=> nil
+```
+
+Maybe that parsing logic ends up looking something like this:
+
+```ruby
+def parse_payment_amount(value, current_balance)
+  amount = BigDecimal(value, exception: false)
+
+  if amount.present? && amount == current_balance
+    [:full_balance, amount]
+  else
+    [:other, amount]
+  end
+end
+```
+
+I return tuples where you either get a verified `:full_balance` amount or you
+get some `:other` amount. The other amount could be `nil` which would trigger a
+downstream validation.
@@ -0,0 +1,51 @@
+# Remove Blank Values From A Hash
+
+Ruby's [`Enumerable`](https://docs.ruby-lang.org/en/master/Enumerable.html) has
+a method `#compact` that will remove `nil` values from `Enumerable` objects like
+hashes and arrays.
+
+```ruby
+> { one: nil, two: 2, three: "" }.compact
+=> {two: 2, three: ""}
+> [nil, 2, ""].compact
+=> [2, ""]
+```
+
+That is often what I want because I'm thinking in terms of values being either
+`nil` or valid value.
+
+Sometimes I want to remove all _blank_ values, not just the `nil` ones. This
+happens in a Rails context like in a controller when dealing with parameters
+from a request -- a form field was left blank (`""`) or no options were picked
+from the multi-select (`[]`).
+
+Rails adds
+[`#compact_blank`](https://api.rubyonrails.org/classes/Enumerable.html#method-i-compact_blank)
+to `Enumerable` to support these cases. Consider that `#compact` corresponds to
+`#nil?` where as `#compact_blank` corresponds to `#blank?`.
+
+Here is the example hash pulled from the docs:
+
+```ruby
+> { a: "", b: 1, c: nil, d: [], e: false, f: true }.compact
+=> {a: "", b: 1, d: [], e: false, f: true}
+> { a: "", b: 1, c: nil, d: [], e: false, f: true }.compact_blank
+=> {b: 1, f: true}
+```
+
+I might want to `compact_blank` a set of search parameters coming from the
+client before passing it into my search service:
+
+```ruby
+class BooksController < BaseController
+  def search
+    @results = SearchService.call(search_params.compact_blank)
+  end
+
+  private
+
+  def search_params
+    params.permit(...)
+  end
+end
+```
@@ -0,0 +1,52 @@
+# Override Output Field Separator
+
+Have you ever seen `$,` as a specific piece of syntax before? I've never
+actually seen any Ruby code that uses this in the wild. It's called the _output
+field separator_. By default it is `nil`. Like other keywords prefixed with `$`
+it is a global variable.
+
+It can be used to control what gets used by default as a separator in variety of
+contexts, like when calling `#join` without any argument.
+
+```ruby
+> [1,2,3].join
+=> "123"
+> $,
+=> nil
+> $, = '~'
+=> "~"
+> [1,2,3].join
+=> "1~2~3"
+```
+
+It is also what gets used when calling `print` with multiple arguments:
+
+```ruby
+> $, = ' |v| '
+=> " |v| "
+> print "hello", "world"
+hello |v| world=> nil
+> $, = "\t"
+=> "\t"
+> print "hello", "world"
+hello	world=> nil
+```
+
+This seems pretty esoteric so far, but might be useful in a one-liner where
+we're doing something like parsing data from a CSV:
+
+```ruby
+# awk-style: pull fields 1 and 3 from a CSV, emit them tab-separated
+ruby -F, -ane 'BEGIN { $, = "\t"; $\ = "\n" }; print $F[0], $F[2]' data.csv
+```
+
+My attention was originally drawn to `$,` by a [post from Chris
+Oliver](https://lnkd.in/p/g6YH5KvQ) which showed this method definition from
+Rails source:
+
+```ruby
+def safe_join(array, sep = $,)
+```
+
+Note that this is [now
+deprecated](https://docs.ruby-lang.org/en/master/language/globals_md.html#deprecated).
@@ -0,0 +1,74 @@
+# Add Check Constraint To Existing Column
+
+To not bury the lede, SQLite does not support adding a check constraint to an
+existing column. Instead, if I want to add one after the fact, I can reconstruct
+the table with the check constraint specified up front and then migrate the
+data.
+
+Here is a `sessions` table that includes `start_time` and `end_time` columns
+that use the `text` data type to represent points in time.
+
+```sql
+sqlite> .schema sessions
+CREATE TABLE sessions (
+        id integer primary key,
+        active integer not null check (active in (0, 1)),
+        project_id integer not null references projects(id) on delete cascade,
+        start_time text not null,
+        end_time text,
+        created_at text not null default (datetime('now')),
+        updated_at text not null default (datetime('now'))
+    );
+CREATE UNIQUE INDEX idx_sessions_single_active
+        on sessions(active)
+        where active = 1;
+```
+
+I want `start_time` and `end_time` to both enforce the shape of the timestamp
+strings with `check` constraints. To do that, I need to start a transaction,
+create a new version of the table with the check constraints, migrate the data,
+rename `sessions` to `sessions_old`, rename `sessions_new` to `sessions` (that's
+the in-place swap), and then commit the transaction. `sessions_old` can be
+dropped later once I feel good about the migration.
+
+```sql
+begin transaction;
+
+create table sessions_new (
+  id integer primary key,
+  active integer not null check (active in (0, 1)),
+  project_id integer not null references projects(id) on delete cascade,
+  start_time text not null,
+  end_time text,
+  created_at text not null default (datetime('now')),
+  updated_at text not null default (datetime('now')),
+  check(start_time is strftime('%Y-%m-%dT%H:%M:%fZ', start_time)),
+  check(end_time is strftime('%Y-%m-%dT%H:%M:%fZ', end_time))
+);
+
+insert into sessions_new (
+  id, active, project_id, start_time, end_time, created_at, updated_at
+)
+select id, active, project_id, start_time, end_time, created_at, updated_at
+from sessions;
+
+alter table sessions rename to sessions_old;
+
+alter table sessions_new rename to sessions;
+
+drop index if exists idx_sessions_single_active; -- on sessions_old
+
+create unique index idx_sessions_single_active
+  on sessions(active)
+  where active = 1;
+
+commit;
+```
+
+Notice that after the table renames I also drop index (now pointing to
+`sessions_old`) and recreate it for the _new_ `sessions` table.
+
+This approach worked well for my situation, but may not be a one-size-fits-all
+solution. Depending on how the database is deployed, the size of the tables, and
+usage, this approach may not scale. Always do a dry-run of database migrations
+like this.
@@ -0,0 +1,62 @@
+# Manage Lightweight Schema Migrations With `user_version`
+
+The [`user_version` pragma](https://sqlite.org/pragma.html#pragma_user_version)
+in SQLite is a bit of persistent state built into the header of a SQLite
+database file.
+
+> The user_version pragma will get or set the value of the user-version integer
+> at offset 60 in the database header. The user-version is an integer that is
+> available to applications to use however they want. SQLite makes no use of the
+> user-version itself.
+
+One use case for `user_version` is as a schema version indicator. It starts at
+`0` which can mean nothing has been migrated yet.
+
+```sql
+sqlite> pragma user_version;
+0
+```
+
+Then for each individual migration that my application applies, I can increment
+the value of `user_version`. At any point in the application lifecycle, it knows
+up to what point schema migrations have been run. If the number of migrations is
+greater than the value of `user_version`, then migrations need to be applied.
+
+Here is some Python code (from
+[`py-vmt`](https://github.com/jbranchaud/py-vmt/blob/e426b3fda92f59c116494a4a4ed992965bac1dba/src/py_vmt/db.py))
+that uses this pattern:
+
+```python
+from sqlite3 import Connection
+
+MIGRATIONS = [
+  # ...
+]
+
+
+# Using the `user_version` pragma in SQLite, this checks if there are any
+# sets of statements in `MIGRATIONS` that have not been run yet. It then
+# executes those and updates `user_version`.
+def migrate(conn: Connection):
+    version = conn.execute("pragma user_version").fetchone()[0]
+    for i, statement in enumerate(MIGRATIONS[version:], start=version):
+        conn.executescript(statement)
+        conn.execute(f"pragma user_version = {i + 1}")
+    conn.commit()
+```
+
+Notice that `user_version` can be updated directly with an assignment:
+
+```sql
+sqlite> pragma user_version = 3;
+sqlite> pragma user_version;
+3
+```
+
+A `select` statement can also be used to access the `user_version` value like
+so:
+
+```sql
+sqlite> select * from pragma_user_version;
+3
+```
@@ -0,0 +1,50 @@
+# Output Query Result In Nicely Formatted Table
+
+When I start a fresh SQLite connection and run a query, all the results are
+squished together in a way that is poorly formatted, hard to read, and missing
+column headers.
+
+```sql
+sqlite> select sessions.id, start_time, end_time, projects.name from sessions join projects on projects.id = sessions.project_id limit 3;
+1|2026-07-26T21:15:50.062936+00:00|2026-07-26T21:53:13.990Z|taco
+2|2026-07-26T21:53:40.019946+00:00|2026-07-26T22:14:09.168Z|TIL
+3|2026-08-02T17:00:16.119169+00:00|2026-08-02T17:30:16.119Z|py-vmt
+```
+
+I can drastically improve the look of this by turning _headers_ on and switching
+to _box_ mode.
+
+```sql
+sqlite> .headers on
+sqlite> .mode box
+sqlite> select sessions.id, start_time, end_time, projects.name from sessions join projects on projects.id = sessions.project_id limit 3;
+┌────┬──────────────────────────────────┬──────────────────────────┬────────┐
+│ id │            start_time            │         end_time         │  name  │
+├────┼──────────────────────────────────┼──────────────────────────┼────────┤
+│ 1  │ 2026-07-26T21:15:50.062936+00:00 │ 2026-07-26T21:53:13.990Z │ taco   │
+│ 2  │ 2026-07-26T21:53:40.019946+00:00 │ 2026-07-26T22:14:09.168Z │ TIL    │
+│ 3  │ 2026-08-02T17:00:16.119169+00:00 │ 2026-08-02T17:30:16.119Z │ py-vmt │
+└────┴──────────────────────────────────┴──────────────────────────┴────────┘
+```
+
+I personally find that much easier on the eyes. It is also a nicer format to
+copy and paste into a post like this or a formatted code block that I'm sharing
+with a colleague.
+
+I can also do this directly from the CLI with a one-liner using `-header` and
+`-box` like so:
+
+```bash
+❯ sqlite3 /Users/lastword/.local/share/vmt/sessions.db -header -box "select sessions.id, start_time, end_time, projects.name from sessions join projects on projects.id = sessions.project_id limit 3"
+┌────┬──────────────────────────────────┬──────────────────────────┬────────┐
+│ id │            start_time            │         end_time         │  name  │
+├────┼──────────────────────────────────┼──────────────────────────┼────────┤
+│ 1  │ 2026-07-26T21:15:50.062936+00:00 │ 2026-07-26T21:53:13.990Z │ taco   │
+│ 2  │ 2026-07-26T21:53:40.019946+00:00 │ 2026-07-26T22:14:09.168Z │ TIL    │
+│ 3  │ 2026-08-02T17:00:16.119169+00:00 │ 2026-08-02T17:30:16.119Z │ py-vmt │
+└────┴──────────────────────────────────┴──────────────────────────┴────────┘
+```
+
+Run the `.help` dot-command from a SQLite prompt for a full listing of these
+commands. See also `sqlite3 --help` from the CLI for usage details about all
+flags.
@@ -0,0 +1,39 @@
+# Connect To Individual Overmind Processes Via tmux
+
+A common `Procfile.dev` in a Rails project might look something like this.
+
+```
+web: bin/rails server -p $PORT
+vite: bin/vite dev
+worker: bundle exec good_job start
+```
+
+Instead of starting up each process that needs to be running for development to
+work, I can instead run a tool that reads the procfile and sets it all up for me
+-- like `overmind`.
+
+```bash
+❯ overmind start -f Procfile.dev
+```
+
+What's cool about `overmind` is that it starts its own `tmux` session and then
+runs each of these processes in its own window.
+
+I can connect to any one of them by name with `overmind connect <name>`. Or I
+can connect to the session defaulting to the first window with just `overmind
+connect`.
+
+If I need to see what is going on with my background jobs, I'll run:
+
+```bash
+❯ overmind connect worker
+```
+
+This behaves like any other tmux session, so I can use my prefix key (`ctrl-z`
+in my case) to access tmux-specific keybindings. Most notably, once I'm done
+looking, I'll want to hit `ctrl-z d` to detach from the session.
+
+I'm already using tmux as my daily driver which means its easy for me to end up
+in a nested tmux session if I connect while already in my development session.
+To help with that, I set up [a forwarding
+prefix](set-up-forwarding-prefix-for-nested-session.md).
@@ -0,0 +1,31 @@
+# Check What Is Inside A Zip File
+
+Before unzipping a file, it can be useful to check the contents. You can learn a
+few things from doing this. First off, are all the files nested under some extra
+top-level directory? How many files are in there? How big are certain files? Etc.
+
+The `-l` flag to `unzip` can help answer these questions.
+
+```bash
+❯ unzip -l zipped-files.zip
+Archive:  zipped-files.zip
+  Length      Date    Time    Name
+---------  ---------- -----   ----
+        0  08-22-2026 08:54   files-to-zip/
+       14  08-22-2026 08:54   files-to-zip/data.json
+       38  08-22-2026 08:53   files-to-zip/notes.md
+       14  08-22-2026 08:53   files-to-zip/hello.txt
+---------                     -------
+       66                     4 files
+```
+
+Notice that the first line is for `files-to-zip/` which tells us that everything
+is nested under the top-level directory that was probably the target of the
+`zip` command. Everything is organized into a table with labeled columns which
+allows us to answer our questions and get a good idea of what is in there.
+
+In my experience, zip files tend to have dozens to hundreds of files, so it is
+even better to pipe the above to a pager like `less`. That way you can search
+and navigate through everything that's in there.
+
+See `man unzip` for more details.
@@ -0,0 +1,28 @@
+# Delete Empty Files With Find
+
+I was discussing a command with a colleague for finding and deleting empty files
+from a given directory. This command involved a series of pipes and I probably
+wouldn't have solved it too differently. I was curious what other ways there
+were of doing such a task, so I asked Claude. The first option it came back with
+taught me about two new-to-me flags that `find` supports.
+
+First is the `-empty` flag which applies a filter on the results to files or
+directories that are empty.
+
+Second is the `-delete` flag which will delete found files and directories.
+
+Those two can be combined to _delete_ any results that are _empty_. Then to
+target just _files_, I can include the `-type f` flag. And if I want to prevent
+it from recursing down some unexpected tree of directories, I could also add in
+`-maxdepth 1`.
+
+To delete all empty files in `specific-directory`, I can run the following:
+
+```bash
+❯ find ./specific-directory -maxdepth 1 -type f -empty -delete
+```
+
+My first thought for a command like this is that I will need to `xargs rm`, so
+it's neat to know about the `-delete` flag in particular.
+
+See `man find` for more details.
@@ -0,0 +1,17 @@
+# Move YouTube Subtitles Out Of The Way
+
+The thing about following along to any kind of programming tutorial on YouTube
+is that current thing being typed, whether into the terminal or an editor, is
+eventually going to be toward the bottom of the screen. If subtitles are
+enabled, which they mostly are now that they can be auto-transcribed, then they
+will appear on top of text being typed.
+
+It's nice having the subtitles, but not if they are constantly covering up the
+current focal point of the video -- the latest line of code or command being
+typed out.
+
+YouTube's web player allows you to grab and drag the subtitles anywhere within
+the player. So, once they start being in the way, you can drag them up to the
+top of the screen and continue to follow along.
+
+[source](https://bsky.app/profile/jbranchaud.bsky.social/post/3mrjd5xktqs2r)
@@ -0,0 +1,58 @@
+# Remove Pages From A PDF
+
+I have a two page PDF. I really only need to second page though. I want to send
+that second page to someone and including the first page with it will only cause
+confusion. I couldn't figure out a sane way to do this on MacOS with either
+Preview or Notes (where the PDF originated).
+
+So, I looked into a tool I'd heard of recently for manipulating PDFs called
+[`qpdf`](https://github.com/qpdf/qpdf).
+
+> qpdf is a command-line tool and C++ library that performs content-preserving
+> transformations on PDF files.
+
+I installed it with `homebrew` and then looked up how "removing" pages works.
+Conceptually, `qpdf` is used to transform a PDF. That means there are a couple
+ways I can think about approaching this.
+
+If I want to preserve every page except for the first page, then I produce a PDF
+that starts at page `2` and goes to the end (`z`) by specifying a range of
+`2-z`.
+
+```bash
+qpdf in.pdf --pages . 2-z -- out.pdf
+```
+
+If I truly only want the 2nd page from the PDF and nothing else, then I can
+produce a PDF made up of _only_ page `2`.
+
+```bash
+qpdf in.pdf --pages . 2 -- out.pdf
+```
+
+Yet another way to think about this is to produce a PDF that includes all pages,
+but then explicitly mark page `1` for exclusion after the range with `x1`.
+
+```bash
+qpdf in.pdf --pages . 1-z,x1 -- out.pdf
+```
+
+And one last way, assuming that I am dealing specifically with a two-page PDF,
+is to use the `rN` syntax where `N` is the number of pages from the back of the
+list. `r1` is the last page, `r2` is the second to last page, and so forth. In a
+two-page PDF, `r2` would then refer to the _first_page.
+
+```bash
+qpdf in.pdf --pages . 1-z,xr2 -- out.pdf
+```
+
+I know I said that was the last, but I'll also throw in that, again assuming a
+two-page PDF, I could produce a PDF made up of only `:even` numbered pages
+relative to the defined range.
+
+```bash
+qpdf in.pdf --pages . 1-z:even -- out.pdf
+```
+
+I'll stick with the first approach, but it was fun to experiment with all the
+other ways this can be done with `qpdf`'s flexible syntax.
@@ -0,0 +1,37 @@
+# Open Current Command Line Prompt In Editor
+
+The [`fc` builtin](/unix/fix-previous-command-with-fc.md) is a nice trick for
+being able to open my default editor with the previous command loaded in. I can
+make quick edits, save and quit, and the updated previous command will populate
+my prompt.
+
+What if I want to work from a blank prompt or I want to edit the command I've
+already started typing out?
+
+[`zsh`](https://github.com/zsh-users/zsh) has a solution for this via a contrib
+widget, `edit-command-line`, that ships with `zsh`. I can load this widget and
+then bind it to something handy like `Ctrl-x Ctrl-e`. To do that, I can run the
+following sequence of commands and even add them to my `zsh` config:
+
+```bash
+autoload -Uz edit-command-line
+zle -N edit-command-line
+bindkey '^X^E' edit-command-line
+```
+
+The `autoload` line creates a function stub with the name `edit-command-line`
+(try `which edit-command-line` after running just that and you'll see what I
+mean). The `zle` line binds the `edit-command-line` widget to the function named
+`edit-command-line` (if they had different names, it would be `zle -N <widget> <function>`). Lastly, I bind `Ctrl-x Ctrl-e` to that `edit-command-line`
+function.
+
+When this runs, it will first try for the `VISUAL` editor, otherwise fallback
+`EDITOR`. I happen to have both of those set to `nvim` in my config. When I hit
+`Ctrl-x Ctrl-e`, `nvim` opens with the current command populated. I can make
+edits, and then when I run `:wq`, the updated command replaces what was there. I
+still have to hit enter to execute it.
+
+I find this most useful when I'm typing a multi-line prompt, especially one
+where I'm trying to get all the details right and I need to copy in values from
+somewhere else. It is also invaluable for writing out [a command like this](/llm/include-a-file-with-message-to-ant.md) that has a real-world cost if
+I accidentally bump enter too soon.
@@ -0,0 +1,61 @@
+# Where And Which Are Whence
+
+I was recently watch some of the commands being run by an AI agent in a coding
+harness. One in particular caught my attention because it wasn't one I was
+familiar with. It was the command `whence`.
+
+The `man zshbuiltins` describe `whence` like so:
+
+> For each name, indicate how it would be interpreted if used as a command name.
+
+I'd describe it myself as a utility for surfacing what a given command name
+corresponds to when it gets executed. For instance, when run against the name
+`cat` in my local setup, I see `bat` which is explained by the alias I set up.
+
+```bash
+❯ whence cat
+bat
+```
+
+There are several flags that can be used with `whence`. The underlying behavior
+for `whence` is what is also used for more familiar commands like `where` and
+`which`.
+
+```
+where [ -wpmsS ] [ -x num ] name ...
+        Equivalent to whence -ca.
+
+which [ -wpamsS ] [ -x num ] name ...
+        Equivalent to whence -c.
+```
+
+I can get a clearer answer that `cat` is an alias with `-c`.
+
+```bash
+❯ whence -c cat
+cat: aliased to bat
+```
+
+And with `-ca`, I can see that `cat` resolves several different times on the
+path with the alias to `bat` taking precedence over all of that.
+
+```bash
+❯ whence -ca cat
+cat: aliased to bat
+/bin/cat
+/bin/cat
+/bin/cat
+/bin/cat
+/bin/cat
+/bin/cat
+```
+
+The other command that uses the underlying `whence` functionality is `type`.
+
+```
+type [ -wfpamsS ] name ...
+        Equivalent to whence -v.
+```
+
+See `man zshbuiltins` for more details. There are more flags which can be used
+to get additional info.
Author	SHA1	Message	Date
jbranchaud	8ee23956d2	Add Connect To Individual Overmind Processes Via tmux as a tmux TIL	2026-08-30 08:19:24 -05:00
jbranchaud	7e4a02e66f	Add Output Query Result In Nicely Formatted Table as a SQLite TIL	2026-08-29 21:05:44 -05:00
jbranchaud	965bade9d4	Add Register SQLite Adapter To Serialize Datetimes as a Python TIL	2026-08-28 13:46:28 -05:00
jbranchaud	2a2a962864	Add Set Permission Mode When Starting Session as a Claude Code TIL	2026-08-27 14:25:05 -05:00
jbranchaud	a1708116bb	Add Add Check Constraint To Existing Column as a SQLite TIL	2026-08-25 20:44:10 -05:00
jbranchaud	85a6dd3644	Add Specify Parameter Name For Click Option as a Python TIL	2026-08-23 18:06:37 -05:00
jbranchaud	453e9edbdf	Add Check What Is Inside A Zip File as a Unix TIL	2026-08-22 09:04:49 -05:00
jbranchaud	bcf957cabd	Add Generate Sample PDFs With ReportLab as a Python TIL	2026-08-21 10:58:37 -05:00
jbranchaud	fe019c3338	Change a couple 'we' to 'I'	2026-08-19 22:56:40 -05:00
jbranchaud	5773961c93	Add Remove Pages From A PDF as a Workflow TIL	2026-08-19 22:52:26 -05:00
jbranchaud	8191e9303b	Add Escape Curly Braces Within Formatted String as a Python TIL	2026-08-19 22:11:53 -05:00
jbranchaud	74e71cf1af	Add Duplicate Current Browser Tab as a Chrome TIL	2026-08-19 12:05:08 -05:00
jbranchaud	a16dcc7760	Add Join A List Of Strings as a Python TIL	2026-08-18 21:42:31 -05:00
jbranchaud	c495194a0d	Add `npm run` Has Some Typo Aliases as a JavaScript TIL	2026-08-17 22:24:05 -05:00
jbranchaud	e3f8446980	Add Initialize A PyTorch Tensor as a Python TIL	2026-08-16 23:11:25 -05:00
jbranchaud	d4ebff5939	Add Override Timezone That All Tests Run In as a Python TIL	2026-08-15 16:06:26 -05:00
jbranchaud	565d6a4680	Add Override Output Field Separator as a Ruby TIL	2026-08-14 21:39:41 -05:00
jbranchaud	3c1ec67965	Add Delete Empty Files With Find as a Unix TIL	2026-08-13 11:19:28 -05:00
jbranchaud	f83b5003de	Add Handle Bad Numerical Amounts With BigDecimal as a Rails TIL	2026-08-12 08:32:38 -05:00
jbranchaud	78c9e92bae	Add Where And Which Are Whence as a Zsh TIL	2026-08-12 01:03:25 -05:00
jbranchaud	2bb402ad40	Add Combine StrEnum With Pydantic For Union Type as a Python TIL	2026-08-10 17:08:27 -05:00
jbranchaud	38b822fb5c	Add Convert Arbitrary Number To Probability With Sigmoid as a Math TIL	2026-08-09 13:17:59 -05:00
jbranchaud	d0496bd8ce	Add Sequence Side-Effect With Rolling Back Inserts as a Postgres TIL	2026-08-08 08:54:19 -05:00
jbranchaud	47b83535b8	Add Annotate Return Type On Enum Function as a Python TIL	2026-08-07 22:27:43 -05:00
jbranchaud	a83030bb6d	Add Filter By Truthy Values With Boolean Function as a JavaScript TIL	2026-08-07 16:44:31 -05:00
jbranchaud	eb419752c4	Add Publish A Package To A Test Env As A Dry Run as a Python TIL	2026-08-06 19:28:35 -05:00
jbranchaud	b484204c2f	Add Remove Blank Values From A Hash as a Rails TIL	2026-08-06 09:25:41 -05:00
jbranchaud	f2de2ce3f4	Add List And Count All Posts In TIL Repo as a Git TIL	2026-08-05 18:26:08 -05:00
jbranchaud	4055ed7a7d	Add Manage Lightweight Schema Migrations With User Version as a SQLite TIL	2026-08-04 17:15:28 -05:00
jbranchaud	0866968be3	Add Mark A Release With An Annotated Tag as a Git TIL	2026-08-02 20:22:22 -05:00
jbranchaud	2e34d6480b	Add Experiment With SQLite Queries In Memory as a Python TIL	2026-08-01 19:10:00 -05:00
jbranchaud	64b464c931	Add Move YouTube Subtitles Out Of The Way as a Workflow TIL	2026-07-31 23:40:54 -05:00
jbranchaud	1a0b02168e	Add Another Way To Mark Keyword-Only Dataclass Fields as a Python TIL	2026-07-30 22:10:21 -05:00
jbranchaud	146274230e	Add Run Scheduled Action To Commit Regular Updates as a GitHub Actions TIL	2026-07-29 18:21:47 -05:00
jbranchaud	300a22f18b	Fix indentation of latest TIL	2026-07-29 01:23:23 -05:00
jbranchaud	1225349978	Add Dedent Common Whitespace From Multiline String as a Python TIL	2026-07-29 01:00:46 -05:00
jbranchaud	0fbf06a2ce	Add Open Current Command Line Prompt In Editor as a Zsh TIL	2026-07-28 17:51:39 -05:00
jbranchaud	2c247f2e20	Add Start Jupyter Notebook With Extra Packages as a Python TIL	2026-07-27 18:12:36 -05:00
jbranchaud	6a67541aa4	Add Include A File With Message To Ant as an LLM TIL	2026-07-27 09:04:16 -05:00
jbranchaud	57188341a6	Add Configure Other Attributes Of Dataclass Field as a Python TIL	2026-07-26 19:12:31 -05:00