@garrytan Garry - you should do a quick video about how you are using it end to end
English
Mike Endale
2.4K posts
Mike Endale
@MikeEndale
I build products that make real impact and invest in interesting ideas. Current: co-founder & vp @blencorp. Previous: Founder of Moxit (acq by @wonderschools)
Washington, DC & Addis Ababa Katılım Haziran 2008
907 Takip Edilen3.2K Takipçiler
@arvidkahl @GeoffreyHuntley At the end of the day, everything is a Ralph loop in the AI world.
English
Isn’t /autoresearch just a fancier name for Ralph loop?
Anyone in the know able to explain the difference?
tobi lutke@tobi
OK, well. I ran /autoresearch on the the liquid codebase. 53% faster combined parse+render time, 61% fewer object allocations. This is probably somewhat overfit, but there are absolutely amazing ideas in this.
English
@garrytan @garrytan look at Cloudflare’s Code Mode MCP implementation. It’s able to work through what would’ve been a million token is just a few thousand tokens.
blog.cloudflare.com/code-mode/
English
MCP sucks honestly
It eats too much context window and you have to toggle it on and off and the auth sucks
I got sick of Claude in Chrome via MCP and vibe coded a CLI wrapper for Playwright tonight in 30 minutes only for my team to tell me Vercel already did it lmao
But it worked 100x better and was like 100LOC as a CLI
Morgan@morganlinton
The cofounder and CTO of Perplexity, @denisyarats just said internally at Perplexity they’re moving away from MCPs and instead using APIs and CLIs 👀
English
@karpathy @Yuchenj_UW I think the Ralph loop works better when driven from a single PRD json so that you can continue to update the file and steer it
English
Codex is a know issue :( It basically don't work with autoresearch sadly, in the way it's set up atm:
github.com/karpathy/autor…
I pung a friend at OpenAI to see if something can be done, e.g. need a /loop equivalent or something like that. More generally, I really dislike the -p + ralph loop pattern of running agents "headless". I want nice, interactive sessions running in tmux so that I can see what they are doing, pitch in, etc.
English
GPT-5.4 xhigh seems bad at following instructions.
Last night I launched two AI research agents running @karpathy’s autoresearch.
Claude Opus 4.6 (high):
> ran for 12+ hours, 118 experiments done, still running
GPT-5.4 xhigh:
> stopped after 6 experiments
> blamed me for “manually interrupting” it
> I interrogated it
> It admitted it made a mistake and stopped the loop itself, despite an explicit LOOP FOREVER instruction in the md file. 💀
English
@thsottiaux 1. option to allow it to run in yolo per conversation.
2. hooks
English
@jolingkent @DarioAmodei @CBSSunday Jo, it was really frustrating listening to the interview. I don’t know why you did not ask him a basic question at the top. x.com/mikeendale/sta…
Mike Endale@MikeEndale
This interview by @DarioAmodei on Pentagon feud is really frustrating. I don’t know why @jolingkent did not simply asked if DoW does any type of surveillance on American citizens? If yes, well, that’s a massive problem. If no, why are is Dario so concerned about it? youtu.be/MPTNHrq_4LU?si…
English
“We are patriots.” Watch more of my exclusive conversation with Anthropic CEO @DarioAmodei tomorrow on @CBSSunday 🌞
CBS News@CBSNews
Anthropic CEO Dario Amodei said in an exclusive interview with CBS News' @jolingkent that the AI company sought to draw “red lines” in the government’s use of its technology because “we believe that crossing those lines is contrary to American values, and we wanted to stand up for American values.” He added: “Disagreeing with the government is the most American thing in the world.”
English
This interview by @DarioAmodei on Pentagon feud is really frustrating. I don’t know why @jolingkent did not simply asked if DoW does any type of surveillance on American citizens?
If yes, well, that’s a massive problem.
If no, why are is Dario so concerned about it?
youtu.be/MPTNHrq_4LU?si…
YouTube
English
The reason why most companies would prefer the @claudeai Max account over API is pure price producibility. Anthropic should just start charging $1000 for unlimited and stop bothering people. Let them use the account as they wish.
There is a demand for it.
English
@mattpocockuk Same. This has been the most reliable process that I have found in almost a year and half of experimentation. Lisa to plan, Ralph to build and test. The models (especially Codex) has been doing an incredible work
English
All the investments I've made in my ralph loops and skills are really paying off now.
I am producing more high-quality code than ever and there are still so many low-hanging fruit to pick.
Super-exciting times to be a dev.
English
@mattzcarey I’m just did a first pass and I don’t know I believe it yet 😀
The API I am working on is roughly 1.1M tokens in full
Will report back 😀
English
This is legitimately a big deal. Has anyone tested it yet? I am spinning it up now. Curious to hear your thoughts
Matt Carey@mattzcarey
Code Mode is all you need, very excited about this direction for MCP blog.cloudflare.com/code-mode-mcp/
English
@caelin_sutch It won’t be too far. Working on a skill to see how well it will do
English
Claude make me FedRamp certified high, make no mistakes
Claude@claudeai
Introducing Claude Code Security, now in limited research preview. It scans codebases for vulnerabilities and suggests targeted software patches for human review, allowing teams to find and fix issues that traditional tools often miss. Learn more: anthropic.com/news/claude-co…
English
Getting accounting setup for my startup.
I use @mercury for banking so thinking about using their integration with @QuickBooks.
Suggestions?
English
@mattpocockuk One thing worth adding to that is to say at the end “use the AskUserQuestion tool”
It greatly improve the interview process
English
Claude Code (or Opus 4.6) feels like it asks you far fewer questions during plan mode
Try:
"Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one."
English
This is such a great model. I hope @alx_ethiopia will follow suit quickly and get some of the capabilities in Ethiopia.
The opportunity to use some of these agent capabilities in health in Ethiopia has a massive upside.
cc: @FMoHealth @GMirafe
Anthropic@AnthropicAI
We've signed an MOU with the Government of Rwanda—the first partnership of its kind in Africa—to bring AI to health, education, and other public sectors. Read more: anthropic.com/news/anthropic…
English
It’s really cool to see @mukundjha, @nitprashant and the rest of @emergentlabs team passing a major milestone. It was great to see the team in action a few months at @ycombinator.
Congratulations to all of you. Incredible feat!
Mukund Jha@mukundjha
Emergent is now at $100M ARR (run rate) We got here in 8 months, making @emergentlabs one of the fastest-ever to do it! That's not all. Today, we're unlocking the next billion builders who will build on the go. 📱
English
@sama Called this would happen a few weeks ago. This is a natural home for @openclaw. Congratulations @steipete - you’ve built magic
x.com/mikeendale/sta…
Mike Endale@MikeEndale
@levelsio I think Anthropic will just build into Claude Cowork, which already has all the harness it needs. I actually think the better home for @openclaw is OpenAI. The ChatGPT desktop app is installed in hundred (maybe more) millions of computers already. This is a natural extension
English
Peter Steinberger is joining OpenAI to drive the next generation of personal agents. He is a genius with a lot of amazing ideas about the future of very smart agents interacting with each other to do very useful things for people. We expect this will quickly become core to our product offerings.
OpenClaw will live in a foundation as an open source project that OpenAI will continue to support. The future is going to be extremely multi-agent and it's important to us to support open source as part of that.
English
@felixrieseberg @jukan05 Felix - one thing I really would love is to have an “incognito” mode similar to Claude. There are a bunch of use cases where users will not want to have certain things to be part of the chat history
English
@jukan05 Thanks for the kind words, please let us know how we can make it better - I’d love to know some of your flows!
English
@ryancarson @clerk Ryan - so they have a testing token bypass if you want to try it. I have a flag in .env to do that on dev.
clerk.com/docs/guides/de…
English
@clerk please please please turn on 'allow agent login' so we can do browser testing with agents easier.
This is way too freaking hard.
give us an agent-first way to do this instead of hacky playwright scripts and praying to the google chrome devtools gods.
all of us need e2e testing and it should be easier.
English