Fahad

20.
@uncensored_ai @asksurplus

prev building @heuristicsai

[Thariq @trq212]

I’ve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined.

I have really loved working at Ant and believe deeply in the mission, and most of all, the idea of a race to the top.

The idea that if we work hard and responsibly, we can raise the bar and inspire others to great outcomes. And if we all push each other to be better, things can be radically good. I will carry the spirit of this with me forever.

My colleagues at Anthropic are the most talented, thoughtful, competent, and earnest people I have ever met. I am glad they are at the wheel and trust them deeply.

To the community on X, thank you for everything. You’ve made this the most fun (and sometimes stressful) job I’ve ever had.

For me, it is time for a bit of a break before diving into something new.

If you want more excellent Claude Code content, you should follow @addyosmani , @lydiahallie, @_catwu, @ClaudeDevs and of course @bcherny who started it all.

[Fahad @fllennn]

@mac_eth @PublicProof yes, worth considering

[Surplus Intelligence @AskSurplus]

https://t.co/ztvp289Lcc

[Fahad @fllennn]

when someone asks me what hockey stick growth looks like, I show them this.
https://t.co/kuc7UI8Cvn rising!

[Fahad @fllennn]

everything is a denominator problem.

make the denominator so big that even at the slightest success rate, the numerator becomes a real number

expected wins = attempts × success rate

success rate depends on a hundred things you can't fully control. attempts are the one thing you can control completely.

take so many attempts that even at a marginal 0.1% success rate, 1000 tries gets you about 1 win on average. and most of the time 1 win is all you need to offset every failure before it.

upsides:

you get the win
you get better at it by doing it 1000 times
you stop caring about rejections

downside:
there'll always be a million reasons not to do something. but the real downside is usually just a "no". the pros far outweighs the cons.

and remember, you miss 100% of the shots you don't take.

[Fahad @fllennn]

@swymbnsl @rhutik_parab just go straight to prod bro🥀

[Fahad @fllennn]

@_Creation22 500K LOC PR btw. Bro one shotted the entire app.

[Fahad @fllennn]

@_Creation22 @uncensored_ai LFG!!!

[Fahad @fllennn]

let him cook

[Fahad @fllennn]

GPT-6 Astra and Fable 5.1 releases were cool, but this gets me going more than anything else.

[OpenAI @OpenAI]

ChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.

- Faster image generation to keep your ideas flowing

- Improved fidelity for more natural, recognizable images

- Consistent details across multiple edits

- Comment-based edits to change only what you want

[Fahad @fllennn]

@menace_codes Switch to @AskSurplus and cut it down by 90%. That'd be more of a flex imo

[Capminal @Capminal]

1/4 Integration Announcement 🟪

@AskSurplus has integrated with Capminal OpenCAP as an inference provider. Now, $CAPU stakers who receiving daily inference credits on OpenCAP can create an API Key and sell their models on Surplus.

Huge thanks to the Surplus team for putting in the work to make this integration happen 🫡

[Surplus Intelligence @AskSurplus]

Built something using Surplus?

Submit your project and we’ll add it to the directory

[Fahad @fllennn]

Built something with @AskSurplus?

We're putting together a Built with Surplus page to highlight the ecosystem.

Reply below with what you've built (and a link), we'd love to feature it.

[Fahad @fllennn]

The vertical AI company to proprietary data flywheel to model training pipeline is becoming very real

What naysayers called "chatgpt wrappers" initially are turning out to be future AI labs. Many more to join lol

[Gabe Pereyra @gabepereyra]

We are hiring for @Harvey’s model training team.

This team will help Harvey expand from the application layer into the model layer and from legal into high end knowledge work more broadly.

We are hiring AI researchers of all seniority, particularly those with experience post-training frontier or open source models.

Our program is centered around large-scale model training, synthetic data generation, long horizon reinforcement learning, and rigorous evaluation in real world deployments. We are scaling-pilled and believe that nothing beats the combination of larger models and better training data.

We’ve been able to generate incredibly realistic legal environments and validated that this allows us to post-train open source models to achieve frontier performance with agents. We plan to scale up these data generation and training efforts significantly across legal to start, and eventually other verticals.

As a researcher, you will have access to thousands of GPUs and unique training data from our product and customer relationships.

Your research will inform Harvey’s product strategy and power AI used for some of the most economically and societally impactful work in the world.

[Fahad @fllennn]

We’re entering an era where almost anyone can build. The cost of turning an idea into a product is collapsing, and technical ability is becoming increasingly accessible.

But building something and being a founder are not the same thing.

Being a founder is more about operating discipline than technical building. It is judgment under uncertainty, allocating scarce resources, making consequential decisions with incomplete information, hiring and leading people, setting priorities, and absorbing responsibility when there is no obvious answer.

The tools required to create are being democratized. The capacity to operate is not.

More people can build companies than ever before. Far fewer can actually run them.

No matter how good AI becomes at building your company for you, the fate of the company depends on the operating discipline of the founder.

[Fahad @fllennn]

Sometimes I just want ChatGPT and Claude to talk to each other, sort it out between themselves, and come back to me when they’ve reached an agreement.

[Fahad @fllennn]

Absolutely no way on earth you need $136M to build this.

[Tanmai Gopal @tanmaigo]

We raised $136M to kill Slack.

Introducing PromptQL: The first AI version of Slack.

Here’s how it works:

[Fahad @fllennn]

Guys, I think Egypt won

[ali @waterloo_intern]

we distilled 2.3M Claude Fable 5 reasoning traces into Qwen3-4B

- 100% self-consistency @ 512 samples
- 0.00 bits output entropy
- zero hallucination variance

turns out the student is not bounded by the teacher.
it also converged on one universal truth.

we open-sourced the model weights👇

[Fahad @fllennn]

we're so back