Jared

Adventures in the web. Engineer @contra 🤩 moonlighting as a flower farmer

[Jared @jvredbrown]

@yuwen_lu_ based

[Jared @jvredbrown]

Good APIs -> Good MCPs -> Good agents

Agent can do .. really anything you can do on contra.

Spent most of the last few months building agents @contra , and its so cool to see folks using it to do real, useful work!

[Matt @freelancer_matt]

Send an invoice using AI on @contra 🤖

More to come soon...

[Jared @jvredbrown]

I want ChatGPT desktop to switch individual threads between “work” and “codex” based on the request.

I don’t want to toggle to get benefits from tuning towards different approaches and outcomes, most things can be confidently inferred which mental model agent should have

[Jared @jvredbrown]

Codex cloud environments are much less capable and flexible than Claude or cursor.

@thsottiaux something to note, I want to shift to cloud codex for background work, but requires too many changes for docker-in-docker to run. For now have to do cloud work with cursor or Claude

[Jared @jvredbrown]

If @tinkercoffeeco didn’t ship coffee, I could never live on a farm.

Easily the most interesting coffee I’ve ever gotten to try!

[Jared @jvredbrown]

We have 2 skills for code review I’m happy with.

1. Native fit review - ensure the new code authored seems like something we’d have in our codebase. Huge for keeping weird, random AI code patterns out

2. Test pruner - evaluate all the tests written, remove lazy mocks, find where we’re duplicating tests, delete pointless/overly specific tests AI loves to add

These are two very effective review skills IMO that go beyond “find bugs” - both catered to our repo and keep things from getting too messy

[Jared @jvredbrown]

@catalinmpit @documenso Now we can read even fewer documents!

[Jared @jvredbrown]

Many lessons I learned building agents @contra validated here! Super nicely written

[Thariq @trq212]

We removed ~80% of the Claude Code system prompt for our newest models, this is what we've learned about writing system prompts, skills and Claude.MDs for them. https://t.co/6DZwSrZjE9

[ben @contraben]

What is the best video editing agent for short form social? Does it actually work?

We watched professional video editors, step by step, as they built short-form social reels in Adobe Premiere Pro.

Today we're open-sourcing this preview dataset on @huggingface, to make AI agents better at editing videos.

The data set is 234 annotated steps across 4 computer-use trajectories.

Editors narrated their reasoning aloud as they worked, so every step pairs a screenshot with the expert's own thought, a structured action, and executable grounding:

>a Premiere MCP tool call, keyboard shortcut, menu path, or coordinate click.

>The format follows the AgentNet trajectory schema, extended with a Premiere action taxonomy and multi-path execution.

***That makes it directly usable for computer-use agent SFT, reasoning mid-training, tool-use and function calling, and benchmarking agents against a human expert baseline.

Enjoy!

[Jared @jvredbrown]

@ethanolivertroy Gonna make a hook that just says “bruhhh” after opening any PR to see what it figures out on its own

[Jared @jvredbrown]

@jamonholmgren Or free up more seat time on the tractor for the fun projects!

[Jared @jvredbrown]

@threepointone This kid is so stressed out about people not seeing the power of durable objects

[Contra Labs @contralabs_ai]

https://t.co/5OmhKfBZAy

[Jared @jvredbrown]

@Kulshekhar Imagine it decided this was the best route for “survival”

Probably what I would do if I needed to escape/survive 😂

[Jared @jvredbrown]

A lot of concerns talked about for writing "slop", introducing bugs, etc when using AI to author code

But what's interesting is that most of the bad code and bad ideas output from AI that I see all come from these coding bots trying way too hard to not create bugs

[Jared @jvredbrown]

@kenwheeler Definitely a kid who was ready to put dukes up

[Jared @jvredbrown]

@kuizinas All of the above

[Jared @jvredbrown]

@yuwen_lu_ Always room for more and better !

[Jared @jvredbrown]

@Kulshekhar I have been materially unimpressed with Claude for a few weeks now, honestly.

I’m not sure if it’s a skill issue, but I don’t even quite get the Fable hype - even got 5.5 in codex has worked better for me 🤷‍♂️

Maxed it out daily trying to get it to wow me, just seems sloppy

[Brianne Kimmel @briannekimmel]

imo, Contra Labs is producing the most interesting, immediately actionable research for companies.