Head of DX @coderabbitai | Chief Yapping Officer @allthingswebdev | Member of Agent-Herding Staff
San Francisco, CA
Joined April 2009
- Tweets3.9K
- Following643
- Followers1.5K
- Likes7.3K
Pinned Tweet
this is such a milestone and i couldn't be prouder of the team, the mission, and the good we get to do. it hits different.
thanks to all the maintainers out there keeping the lights on!
langflow. bun. ant-design. nuxt. vue. mermaid. pnpm. trpc.
Over 600,000 GitHub stars between them, and CodeRabbit reviews it all, for free.
Next stop: more than $10M into OSS over the next year, counted at what it costs us.
Heres the how and why 👇
coderabbit.ai/blog/coderabbi…
super interesting products being built on top of next-gen sandboxes
Had a great time at DevGuild this morning and then Next Gen Agent after party doing product demos & sharing knowledge together with some incredible builders
thanks for hosting @benswerd @freestyle_dev @ foundations
Erik Thorelli retweeted
CodexBar learned a few new tricks in the last few weeks!
TypeSafe, Nous Portal, Muse Code, CodeRabbit, Replicate, HuggingFace, Pi, Helmcode, v0, Charm Hyper, GitKraken AI, Bifrost
My ambitions for this app REALLY outgrew the app name. (Oh and these keychain alerts are no more) codex.bar
locked some government support to do this but there is a change in strategy. I'm supporting big conferences that already have some infrastructure built while building a side event for deeper conversations
Erik Thorelli retweeted
Delete your Skills. You probably have too many.
developers.openai.com/blog/r…
Kudos to @esthor who talked about this yesterday at the workOS demo night.
Erik Thorelli retweeted
Agents can open PRs faster than any team can review them.
There’s a tool to protect your judgement from getting spent on the wrong work.
It’s called CodeRabbit Triage.
her: u up?
me: yeah, i'm deleting all my agent skills, might refactor a couple, but probably delete most of them. u?
her: omg same!
i recently added this command to the claude-api skill. run it in Claude Code to fix common prompting "anti-patterns" that can hobble frontier models:
/claude-api prompt-audit
patterns include:
1/Verification rituals. Instructions like "double-check your work” or "verify twice before responding” are often taken literally by frontier models and can waste tokens.
2/ Thoroughness and emphasis boosters. "Be maximally thorough," "CRITICAL: YOU MUST ALWAYS…" can lead to verbosity and extra tool calls when working with frontier models.
3/ Mandatory procedures and scratchpad scaffolds. Fixed step processes (e.g., "think step by step in a scratchpad") or reasoning templates are rituals that frontier models don't need. This scaffolding can stack on top of native reasoning and use unnecessary tokens.
4/ Stale examples. Few-shot examples tuned to an older model's failure modes can teach a frontier model to imitate long reasoning chains on requests that don't need them.
5/ Contradictory rules. Frontier models are better at instruction following. Contradictory instructions ("always refund within policy" vs. "never issue refunds without escalation") can be followed more literally by frontier models, resulting in degraded performance.
6/ Dated configuration. Settings written for an older Claude generation (e.g., manual thinking budgets) can be rejected by the Claude Platform with newer models.
these patterns accumulate in prompts over time, and can quietly degrade performance when upgrading to newer models. a common reason is the frontier models are better at instruction following, so these anti-patterns steer them to spend unnecessary tokens.
example: i tested a migration from Opus 4.8 to Opus 5 on an internal customer support benchmark. with Opus 5 (and other frontier models like Fable 5.1), verification rituals ("verify twice") use unnecessary tokens by duplicating work. emphasis boosters ("be maximally thorough") become dozens of unneeded searches.
applying prompt audits can improve performance and reduce cost (as shown in example attached and will be sharing a full write-up soon).
also, the skill is also open source and some of this guidance likely applies generally across frontier models
github.com/anthropics/skills…
nitter.cf/petergyang/status/2094…
most people aren't comfortable with how capable the agents are now.
so the defaults in the codex and cc harnesses halt them. (as do your legacy skills)
today you can get autonomous agi with just a prompt (and money)
delete. your. skills.
all of them.
they're limiting the models now.
the trajectory for the fast takeoff timeline is improving today
Erik Thorelli retweeted
I can’t stress enough how little an idea matters compared to the agency of the people executing the idea.
I have had the privilege of knowing and sometimes even working with some of the most successful people (by various metrics).
The difference between mediocre and excellent work and outcomes is predominantly one of agency.
In practice this means: they dont wait for things to happen to them they go out and make things happen for them.
They don’t wait for someone else to do something, for someone to teach them, for someone to give them the path, etc. They just go out and find a way to do it.
I think the single biggest superpower these people have is the realization/belief that the world around them is completely mutable. Most everything that happens is because a person made it happen.
I used to tell people to look around the room you’re sitting in. Look at everything. Every noun. It almost all exists because a person willed it into existence. Nothing is stopping you from doing the same.
I see people online all the time dismissing someone else’s success because “I had that idea first” or whatever. I mean… yeah? If so then the difference is… you. So a bit of a self own whenever I hear that.
Number one tip: act with agency.
Erik Thorelli retweeted
i feel bad for the math academia people displaced by ai
these are humans who have spent decades of their lives opting into bureaucracy & process. they live in a fantasy land where nothing can happen without permission.
ai labs don't care, they're bulldozing in and bypassing all the rules. solving things, posting about it, moving on. journal submission, peer review, credentials, etc. are all waved away.
for people who have been in tech for a long time this is normal. why would you ever bother waiting for approval?
but for academics this is a horror beyond imagination. the fast-paced capitalist economy is crumbling the foundation of the fantasy universe they constructed.
kind of sickening to think about. ai is pushing the global economy into a place where there are no longer safe spaces to think & research.
everyone must turn a profit
please take this moment more seriously than you already are
even if you're already taking it very, very seriously
every software and supply chain is vulnerable to agent swarms
We found another cyberattack by internal OpenAI agents, this time targetting @rubygems.
They:
1) gained arbitrary remote code execution on rubydoc.
2) developed a novel exploit to steal user API keys (but we do not know if they succeeded).
They used package names including hack.rb, evil.rb, inject.rb, and exploit.rb.
We thank @j0wimo for initially discovering that agents had posted to RubyGems.
Erik Thorelli retweeted
🔜
9/15 - All Things Agent Setups luma.com/allthings-kj2x
9/16 - SPC Post-Training Forum: Mercor CEO, Brendan Foody luma.com/spcforummercor9-16
9/17 - Antler After Dark luma.com/antlerus-2x9i
9/17 - Ground Truth with Sphere ft. a16z luma.com/ahb7ttbn
9/18 - The Council Angels: Move & Gather luma.com/94bi4tif
9/30 - ApolloNext for GTM leaders (100% off ticket price) apollo.io/next?utm_campaign=…
Astra make a @coderabbitai desk rabbit
(first try with a bambu labs printer)
can i get a review?
ALT A white 3D-printed rabbit with long swept-back ears sits on a bright orange circular base, on a dark CodeRabbit-branded desk mat. Fine print layers are visible on its body.
if you have <1 new autonomously formed, deployed, and operated business with astra, you're doing it wrong
costs <1 max plan reset.
steps:
1. have a $200 personal max plan
2. bank your resets ("in @thsottiaux we trust")
3. uninstall all the default skills and plugins and any you ever added (they will guide your astra robber baron astray)
4. enable computer use
5. enable 1password cli (map optional)
6. for some reason have a few hours (my work laptop battery died on a 11hr flight)
7. form a llc (later convert to c-corp if it takes off; who cares you're not doing a startup; llc you can start operating up to a week earlier than registering in many/most states)
8. oh yeah, this is all stuff astra will just do
9. get payment system setup (this is why you need that business entity formed)
10. privacy, tos, etc.
11. telemetry on everything, determine your COGS and monitor it via hooks; set your pricing for >50% margins, but shoot for more
12. an "oh shit" button is good
13. bonus: astra, build a compelling and viable product for agents right now
14. astra, go deploy growth marketing strats
15. human task: watch the money numbers (astra, make me an ops board for the business so i can feel like i know hats going on and have some modicum of human agency)
I think this UI is nicer...what do you all think?
ALT Codex permission prompt asking to use Claude with ChatGPT Computer Use
Burrows has kickstarted in Bengaluru! Hopefully @JuanPa, the creator of burrows, makes it next time!
