dad • rock climber • photographer • eng @openai
Washington, USA
Joined May 2008
- Tweets4.2K
- Following149
- Followers15.8K
- Likes1.2K
Come help me/us make a better experience for over a billion users!
btw, am now leading ChatGPT/Codex Web + Desktop infra group
if you’ve already made other developers meaningfully faster (platform, tooling, reliability, DX), the opportunity is to do the same work at AGI scale
hit me up if interested
6
30
3,956
Brent Traut retweeted
I spent 20,000 dollars in API credits having different AI models play each other at Blood on the Clocktower (a Mafia or Werewolf type social deception game).
The results were fascinating. Models have gotten much better over the course of this year at creative on-the-spot scheming and bluffing, and also at reasoning as a team to break apart these underground conspiracies.
I go through some of the most standout games these AIs played in this video. The new models are getting pretty good. If my friends pulled off some of these moves while we were playing, I’d give them a fist bump.
Fwiw, OpenAI’s newer models are much better than Anthropic’s at this game.
You can see the full benchmark at the website, and I’ve also made it so that you and your friends can play again the AIs: clocktowerbench.com
Took a while, but I think I managed to squash all the places we were leaking memory for long-running turns. You should be able to /goal as long as you like now!
Side chats in ChatGPT no longer expire or disappear when you restart the app!
When I joined OpenAI, I found this to be one of the biggest pain points for side chat users, and as of today, it's fixed. Never lose data just because you updated, restarted, etc.
For users who prefer ephemeral side chats: You can manually delete/archive, they get archived/deleted with their parent threads, and you can always just start a new thread and ask it to clean up old ones.
This is one of the best interviews I’ve heard in a while. Very inspiring for the next phase of software development.
This is f*cking insane. SpaceXAI engineer (ex-Cursor):
"99% of people run one GrokBot for everything
I run 70+ bots, and I only talk to my Chief of Staff"
In just 8 minutes, she shows how to turn one GrokBot into a team that runs your email, calendar, Slack and code 24/7
One GrokBot → Specialist Bots → Chief of Staff → Cloud Agents → A Team That Never Clocks Out
A Specialist Bot does one job and keeps its own memory
A Chief of Staff bot watches your Slack and kicks off the work before you ask
Anyone can chat with one bot, almost nobody has a team of them doing the work
Worth more than any $4,000 AI agent course
Bookmark it, watch it today
Then read how to give your GrokBot a real team in the harness engineering guide below ↓
2
71
12,118
Brent Traut retweeted
hi all, two things:
1. usage limits are back to being double – your feedback worked
2. we today kicked off an intense sprint to dramatically improve the app simplicity and performance. i will oversee this. more to come.
now, send me your complaints
Day 25/
We are doubling all usage limits for all Pro 100/200/500 tiers and Business Premium, effective today.
We heard your kind feedback and won’t be reducing usage limits for Pro 200, so your Work and Codex usage will stay the same. We’re also going to continue offering extended limits for dots for now, and usage of your dot will still continue to be separate from your default usage.
And we'll of course also keep shipping better models and other efficiency improvements that increase your effective usage over time.
Pro Codex tip:
1. Get yourself a skill that babysits PRs. It should address feedback, handle merge conflicts, and navigate CI failures. It should alert you if it actually needs you. See openai/codex repo for a good one.
2. Create a new custom section in ChatGPT app called "🚢 Outbound". Drag some threads into it.
3. Create a new Codex thread and name it "🤖 PR Shepherd". Ask it to create a schedule to run every 15 mins. For every thread in the Outbound section, it should run the babysitter skill until the PR is green, then merge it, and finally archive the thread.
When I'm done making a change in some thread, I ask for review from my team, chuck the thread in my Outbound section, and walk away.
I built a fun new project last week with Dots and Sites. It's a website that tracks my entire wardrobe, suggests outfits for me to wear (to break me out of my blue jeans comfort zone), and also adds suggestions for items I can buy to expand.
I told ChatGPT to look through my purchase history and grab photos online of all the clothes I bought. That got most items, but I also took photos of missing items (bad photos in a poorly-lit closet) and I had ChatGPT just grab them out of my Photo Library via computer use. It then generated catalog-style images of everything with clean backgrounds and color correction while preserving logos and scuff marks and everything.
Every day, my dot adds a few new outfits to it, all using image gen. Some are combinations of clothes I own and some link out to various websites for me to buy stuff. All of them use my likeness. I can hide the bad ones, favorite the good ones, and log the outfits I actually wear.
We truly live in a world where your imagination (and a pro sub) is all you need to build incredible things.
wardrobe.btraut.com
1
12
2,040
I had a 30 minute drive tonight, and I used it to talk to ChatGPT (voice mode) about the midterm ballot. It walked through pros/cons of each initiative, dug into endorsements, and let me ask a bunch of stupid questions without judging. 100% recommended. Vote!
2
12
1,993
Y'all are building with human in the loop too often.
If your SDLC has humans in the loop, it's not going to age well. Models aren't perfect, but that doesn't mean you should give up and do things yourself; improving the loop automation is extremely leveraged work.
AI can trivially read and understand application code, so that becomes the source of truth. What purpose then, if any, do unit tests provide?
Every time I make a change, AI just changes all the unit tests anyway, so there's no safety. It's all overhead.