Utku

working on the last “ai harness” you’ll ever use.

launching it really soon!

shipped 30 apps since 2020

[Utku @fromutku]

@DylanoA4 I couldn't "move on" after reading it for the first time. great line!

[Utku @fromutku]

soon™

[Utku @fromutku]

@VictorTaelin Have you tried giving multiple examples and guidelines in the system prompt? I'd also use hooks to remind the agent here and there.

[Utku @fromutku]

@_can1357 Nice attempt though! I tried configuring my modem yesterday with Claude, it refused to continue after trying the default password…

[Utku @fromutku]

I think you should take "harness" literally. Models are generalists by nature. Yes, it works, but giving a model a set of tools optimized for the task makes the operator's life much easier.

I would also argue that coding is baked into almost every LLM these days, so not seeing anything there makes perfect sense.

[Utku @fromutku]

space whale

[Arthur MacWaters @ArthurMacwaters]

It is absolutely mind-bending that this is a real photo of the recovered Starship

[Utku @fromutku]

I said this, and right after @Bot got good models, Claude started giving API credits. I might have 40 views on my tweets, but nobody can tell me I didn't cause this.

[Utku @fromutku]

OpenAI still mogs with how much they give you as a company. Claude and Grok is not even close.

[Utku @fromutku]

Or enable our subscription to work everywhere and include this amount.

Thanks, but OpenAI and Grok do it wayyy better…

[ClaudeDevs @ClaudeDevs]

We’re rolling out monthly Claude Platform API credits for Max and Team plans:

Max 5x: $100
Max 20x: $200
Team: up to $500, pooled

Works on any model, including Haiku 5.5, in your code or third-party harnesses.

How it works and full terms: https://t.co/qmWL6xVGwr

[Utku @fromutku]

@elshayib_ price matched too! sounds like an easy switch for luna workflows.

[Utku @fromutku]

Impressive feature, but why the useless cinematic announcement video?

[OpenAI @OpenAI]

GPT-6 and Intelligent UI, now rolling out in ChatGPT for everyone.

Intelligent UI in ChatGPT delivers fast, interactive answers that make everyday questions more visual, complex topics easier to grasp, and interactive tools for your task available on the spot.

[Utku @fromutku]

they did the thing

[Claude @claudeai]

One more thing: we’re halving the price of cache reads on Claude Sonnet 5.5, to $0.10 per million tokens. That makes Sonnet 5.5 around 20% cheaper to run on most long-running work.

[Utku @fromutku]

@CompleteSkeptic :)

[Utku @fromutku]

@paulg it's ruby right? i love it when watch makers do this

[Utku @fromutku]

@creativedash magic

[Utku @fromutku]

@yeemio usage you get per week in token terms, everything else is meh.

[Utku @fromutku]

OpenAI still mogs with how much they give you as a company. Claude and Grok is not even close.

[Tibo @thsottiaux]

We shipped four things that were deemed good to great and some math proofs, but the vote is clear and the community demands a reset. I did calibrate it and it *seems* that the game is rigged in reset's favor, but such are the rules at the moment.

Therefore ... the reset has been processed. Enjoy!

[Utku @fromutku]

@theo bloated framework :D

[Utku @fromutku]

cats are so back

[Mistral AI @MistralAI]

Meet Mistral Large 4, aka Le Chonk.

• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.

Open weights release end of October.

[Utku @fromutku]

I think I’m the only idiot who still reads the code agents write.

yes, in many cases they’re really good. But once you mix multiple agents into a complex repo, things slowly start getting messed up.

models naturally take the shortest path to getting the task “done”, and over time that creates slop, inconsistencies, weird abstractions, duplicated logic, etc.

if you’re building something you actually care about, I still recommend reading your code.