Jakob // theCodeDrift

Now that we have pinned posts, it's time for a proper introduction. I code (lots of TypeScript, a bit of Rust, even some PHP), talk music and concerts, and occasionally take mid photos. Come for one, stay for the others. I used to hide my interests. We're not doing that anymore.

A few hours left on DevHunt. If you've tried Taskless, an upvote with your GitHub account helps, and so does a question in the comments. devhunt.org/tool/taskless

Setup is one npx command. It finds Claude Code, Cursor, OpenCode or Codex in your repo, lists every permission change it wants to make, and asks once. Then tell your agent to set up Taskless. Questions welcome on DevHunt: devhunt.org/tool/taskless

If your agent wrote the code and you cleaned it up before opening the PR, the correction never shows up in review. It's sitting in your commit history. Taskless reads those commits too. On DevHunt today: devhunt.org/tool/taskless

We pointed Taskless at getsentry/sentry. It read 9,465 review comments across 4,540 merged PRs and wrote 10 rules. Those rules find 253 fixes in the code today. On DevHunt today: devhunt.org/tool/taskless

🚀 Taskless is live on DevHunt today It turns the corrections you keep making to your agent into rules that run locally and in CI. Try it, upvote it, ask me anything in the comments. I'm there all day. devhunt.org/tool/taskless

I don’t need a habit journal, but I do feel like this is the form factor that could take on those little notebooks

What breaks when AI agents touch your data? CL Kao at @datarecce.bsky.social: agents are very confident, and the first result looks great. Keeping that result maintainable is the hard part. Full episode: www.taskless.io/blog/cl-kao-...

The urge to just put in the value “4”…

Models will sometimes decide to force push. @cmgriffing.bsky.social built GitBench to grade models on the git commands they want to run, before any of them touch a repo. www.taskless.io/blog/chris-g...

CL Kao at @datarecce.bsky.social lets the agent prototype first, then makes it start over with everything it learned. The second build starts with the edge cases already known. Full episode: www.taskless.io/blog/cl-kao-...

I'm giving a video talk Dec 7 at AI Coding Summit Berlin (@gitnation.bsky.social): Static and Runtime Enforcement for Agentic Code. I don't think it's the agent's job to police itself. Free through my badge: gitnation.com/badges/ai-co... #AICodingSummit

⛵ seamux 0.5 is out - drive Cursor's agent CLI from the board, next to Claude Code and Codex - the memory pressure warning is live - Fork sends a tangent to its own worktree npx seamux or github.com/thecodedrift...

"People who don't do front end a lot think front end is solved." @cmgriffing.bsky.social on AI output: you catch every mistake on a subject you know, and take the next page verbatim. www.taskless.io/blog/chris-g...

⛵ Coming in the next seamux release: a memory pressure warning OSX reports memory usage. We can warn you if you're agent-ing too close to the sun. Once it goes from orange to red, we'll begin asking "are you sure?" so osx doesn't auto-close your work. github.com/thecodedrift...

Every accept you click on an agent's request is trust you never wrote down. I talked with CL Kao at @datarecce.bsky.social about when a human should get involved. Full episode: www.taskless.io/blog/cl-kao-...

@cmgriffing.bsky.social builds everything live on Twitch. His project, Messijo, was already in flight when agent output got good. New episode of The Task at Hand. www.taskless.io/blog/chris-g...

I've been trying @codedrift.social's taskless.io on VoxPort. It went through my PR reviews, fix commits and project docs and wrote 6 rules with tests, and you can add your own. First scan flagged 34 spots that had drifted. Try it out with one of your repos. tskl.es/r/6na5TYXbDa...

🚀 Taskless CLI 0.13.0 Setup is one command and one question. Your agent writes the rules and publishes a Taskless report: what each rule finds in your code today, and what that saves in review. taskless.io/blog/taskles...

"Oh no, that's not what I meant." CL Kao at @datarecce.bsky.social: correctness in data is subjective until someone defines it. Recce makes it checkable. Full episode: www.taskless.io/blog/cl-kao-...

⛵ seamux 0.4 is out - project colors sync across every browser and device - Reload conversation picks up a new MCP server without a manual resume - better handling for messages that fail to send to cmux npx seamux or github.com/thecodedrift...

CL Kao at @datarecce.bsky.social on the first thing teams see in CI: the blast radius of a change, column by column. Full episode: www.taskless.io/blog/cl-kao-...

Harry Solovay posted six coding traits for @effect-ts.bsky.social, so I turned them into Taskless rules. Determinism narrows what's left to one call before a decision model sees it. taskless.io/blog/determi...

CL Kao at @datarecce.bsky.social on agents: very capable interns, probably ADHD, no taste, very smart. 2 days of thrill. Then the exhaustion of switching between 10 things. Full episode: www.taskless.io/blog/cl-kao-...

Charlie Marsh posted a thread on ty's new redundant-condition check, and we wanted to know how much of it a generic tool could cover. Taskless rules caught 13 of the thread's 15 examples. The last two need real type inference. taskless.io/blog/taskles...

"...and the use case of creating many throwaway sessions is rare enough and work-around-able enough that it's not presently worth the time to do that refactor" Me with a hidden chrome window via Firebot getting spawned every 2s like clockwork when I'm live... github.com/electron/ele...

CL Kao at @datarecce.bsky.social asked an agent for a complex rebase. After a couple of tries it said it would port the change instead. Full episode: www.taskless.io/blog/cl-kao-...

⛵ seamux 0.3 has themes - start from two colors. seamux keeps their hue and its own lightness, so text clears WCAG AA on any hue - one curl to /theme/set switches every board Thanks @crutchcorn.dev for the WCAG chat that inspired it github.com/thecodedrift...

🚀 Taskless CLI 0.12.0 Every rule Taskless issues is verified before it runs. Restore, roll back and list a rule's revisions. check tells you when it ran rules unverified. taskless.io/blog/taskles...

CL Kao at @datarecce.bsky.social asked his agent to test a markdown reader. It started building a terminal emulator instead of using tmux. "No. Stop." Full episode: www.taskless.io/blog/cl-kao-...

Your agent wrote the migration and the diff looks fine. What did it do to the rows downstream? CL Kao at @datarecce.bsky.social on data review and agents. www.taskless.io/blog/cl-kao-...

So fun fact, raccoons can't see into the red light spectrum. They lack the necessary cones to do so, as most mammals do. Also it's @codedrift.social s Birthday! Happy Mean Girls day too!

Sometimes, you get a silly idea mid coffee-sip and immediately stick something cool in the nightly build

seamux 0.2.2: chats that die on 529s or a sleeping laptop now pick themselves back up. Plus minimizable cards and a clear fix-it page when cmux is unreachable.

Using today’s free will to choose unicorn v narwhal socks

"It's single player to multiplayer." Brian Douglas @bizza.pizza: you can play guitar alone and sound amazing. Bring in other players and it's a different thing. Same with agents running in isolation next to each other. www.taskless.io/blog/brian-d...

⛵ seamux - web UI for cmux with support for claude/codex - secure tunnels as a first-class behavior - localhost w/ http basic support - run with "npx seamux" I'd rather this be out in the world than hidden in a private repo. MIT Licensed github.com/thecodedrift...

Isolation, observability, memory. Brian Douglas @bizza.pizza names three hard problems for production agents. Prompts and models aren't on the list. Isolation is handled. Memory is the one nobody uses. www.taskless.io/blog/brian-d...

🚀 Taskless CLI 0.11.3 Run a single rule with check --rule <id>. Rule ids are unique across engines now. Plus a pile of bugfixes. Bonus: hitting problems? Tell your agent to give Taskless feedback and it writes the report for you. taskless.io/blog/taskles...

Running the CLI migrates your .taskless/ layout. That covers the directory. A rule whose matching changed under a new engine version comes through untouched. taskless update walks a ledger and names what changed for the rules you have. 0.11.2 makes init tell you to run it.

Fork a recorded session. Re-run it against a different model at a different reasoning effort. Sixty times. Brian Douglas @bizza.pizza calls it dreams. You come out with a benchmark: this model, this effort. Go. www.taskless.io/blog/brian-d...

Tire shop: *makes several calls confirming the tires are here for the appointment* I arrive Tire shop: We lost two of your tires you ordered, can we make a second appointment?

Agents write bad frontmatter, usually while editing a file for some other reason. One malformed file aborted the whole Vale run, so taskless check reported [], and [] reads like a clean pass. 0.11.2 fixes that along with a dozen other little things. taskless.io/blog/taskles...

Look, I'm late to the party, but the Auto-Group Tabs extension for Chrome has made the left-hand tab rail 100x better. Automatically ungroups them too if you leave the URL matcher. chromewebstore.google.com/detail/auto-...

"Now it's Claude's job. How did we get here? But we got here slowly." Brian Douglas @bizza.pizza on the line from in-browser editors to Codespaces to handing the work to a model that doesn't live on your laptop. www.taskless.io/blog/brian-d...

Hello oxlint - Taskless 0.11.3 will detect oxlint and offer to help write an oxlint rule for you if it can't find one in the oxlint's vast directory. Markdown files are nice, but enforcement is better. <3 What linter(s) should join oxlint in the next release?

You brainstorm a hypothetical customer list with your agent. Ten minutes later it's telling you about the government contracts you already have. Brian Douglas @bizza.pizza on a made-up thing becoming history. www.taskless.io/blog/brian-d...

"TDD is actually saving the bacon." Brian Douglas @bizza.pizza: if the agent has to pass tests and lint before it commits, the output gets better whether or not the person prompting knows what good looks like. www.taskless.io/blog/brian-d...

Networking ritual: scan each other's LinkedIn, promise to stay in touch, then enjoy the awkward silence where one of us forgets to open LinkedIn for 3 weeks. Skipping the extra steps. I just give people an email that tags how we met so it doesn’t get filtered as spam & stick it on an NFC.

The cats have Giardia because of course they do. Off to add disposable litter boxes to my shopping list.

eslint and ruff settled the style of the code we write, but neither checked the words in comments and prose. Taskless 0.11 runs Vale beside ast-grep, so "the docs always say sign in" becomes a check that fails. www.taskless.io/blog/taskles...