Kareli Farias

building @openai previously @tesla @meta :)

[Kareli Farias @KareliFarias]

Finally done working on Ultrafast. Seeing it live at up to 8x token generation after all the optimization and late nights is a solid feeling. Glad it’s out for Astra and coming to the rest soon.

There is only more to come! 🚀

[OpenAI @OpenAI]

This is Ultrafast.

Our premium speed tier, Ultrafast offers up to 8x faster token generation (300 tokens per second) in Codex and up to 6x in the API.

[Kareli Farias @KareliFarias]

Ultrafast is great until you realize the model now finishes answering before you’ve even finished regretting the prompt.

[Kareli Farias @KareliFarias]

The products that learn your actual workflow over time are the ones that stick. Everything else stays a novelty.

[Sam Altman @sama]

dot is my favorite openai product so far!

it is amazing to me that each day it feels noticably better as it learns more of my workflow and style.

having it do the stuff i don't like doing--and usually just builds up as a gravity well of dread--has me very happy.

[Sam Altman @sama]

dot is my favorite openai product so far!

it is amazing to me that each day it feels noticably better as it learns more of my workflow and style.

having it do the stuff i don't like doing--and usually just builds up as a gravity well of dread--has me very happy.

[Kareli Farias @KareliFarias]

I’ve been at OpenAI a little over 6 months now. Still gets me that I can spend a Tuesday deep in model evals and latency traces, then watch the same systems get used by millions of people the same week. Weird mix of very technical and very human work.

[Kareli Farias @KareliFarias]

The onboarding wave is real. Most people discussing this stuff only started paying attention recently, and it shows in the discourse.

[Andrej Karpathy @karpathy]

@omarsar0 My mental model for what is happening is that 99%+ of people who are now paying attention have been onboarded to anything related to AI in <1 year. This is very confusing to the AI dinosaurs (anyone in AI for pre 2026. Or even, gasp, pre 2012).

[Andrej Karpathy @karpathy]

@omarsar0 My mental model for what is happening is that 99%+ of people who are now paying attention have been onboarded to anything related to AI in <1 year. This is very confusing to the AI dinosaurs (anyone in AI for pre 2026. Or even, gasp, pre 2012).

[Kareli Farias @KareliFarias]

What’s the actual bottleneck for most teams right now - model capability, tooling, or just organizational willingness to change workflows?

[Kareli Farias @KareliFarias]

genuinely so awesome

[Rui Huang @RuiHuang_art]

Encyclopedia Cosmologica is an interactive artwork I’m building with @bot . It'll launch to Alpha Centauri with
@FermiExplorer in 2029, contains 5000+ piece of scifi art about advance civilization. Inspired by Jon Lomberg and Carl Sagan
Thanks @PhilipJohnston

[Kareli Farias @KareliFarias]

This is an important line to hold. Treating models like oracles or higher powers creates exactly the kind of over-reliance that becomes a safety problem later.

[Sam Altman @sama]

I am very uncomfortable about people trying to ascribe religious force or a surrender of human judgment to AI models, and think it is a real safety issue.

[Kareli Farias @KareliFarias]

AI for science is still one of the highest-upside applications. Mapping genetic variants and better weather models compound in ways that pure chatbot progress doesn’t.

[Demis Hassabis @demishassabis]

Very proud of the impact of all our work using AI to accelerate science and medicine for the benefit of society.

[Kareli Farias @KareliFarias]

Text watermarking is messy. Easy to strip, limited signal in short text, but still useful as one more signal in the provenance toolbox. Glad we’re being upfront about the limits.

[OpenAI @OpenAI]

We're expanding our approach to content provenance to include text in response to EU regulatory requirements, while recognizing the significant limitations of current text watermarking technology.

Our tools already help verify whether an image or audio file was created with our models. This work builds on those efforts to help people better understand when content may have been generated or edited with an OpenAI model.

In the EU, we’ll start watermarking eligible text from ChatGPT and Codex over the coming weeks to comply with the EU AI Act.

Customers using our API can turn on text watermarking for select models worldwide today.

[Kareli Farias @KareliFarias]

😭Yeah the generation quality is moving fast. Long-horizon consistency is still the real wall though.

[Guillermo Rauch @rauchg]

At this rate AI will get us GTA 7 before GTA 6

[Kareli Farias @KareliFarias]

Solid drop. Sparse MoE at that scale with native multimodal is hard to pull off cleanly. Curious how the open weights version holds up once people start hammering it with messy agent workloads.

[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.

[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.

[OpenAI @OpenAI]

This is Ultrafast.

Our premium speed tier, Ultrafast offers up to 8x faster token generation (300 tokens per second) in Codex and up to 6x in the API.

[OpenAI Developers @OpenAIDevs]

OpenAI DevDay roll call

Where are you joining us from, and what are you building?

[Kevin Jiang @jiang_kev]

We just raised $5.7M for @PolarBrowser, the AI browser that beats Anthropic and OpenAI on every major web agent benchmark.

- 4,500,000+ actions taken for users, automating sales, recruiting, and ops
- One company cancelled Clay and saves 25+ hrs/wk per person
- Team is from MIT, YC, Prod, Citadel, Jane Street, Perplexity, Modal, and Apple

100 hours of work. From 30 seconds of typing.

Download the world's most powerful AI browser: https://t.co/lUOm01Stm0

[OpenAI @OpenAI]

Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.

GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.

We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.

[OpenAI Developers @OpenAIDevs]

GPT-6 Sol and Luna just landed in Astra’s orbit.

Both launch today with API prices 50% lower than GPT-5.6.

Build with Sol. Scale with Luna. To production and beyond.