Nicolas Loterstein

Making AI do the parts of work nobody wants to. @deel .Views: mine, sometimes the model's.

[Nicolas Loterstein @NicoLoter]

So true...

[Guillermo Rauch @rauchg]

Anyone that works in AI is working the hardest they've ever worked in their lives. On the surface it's somewhat ironic (AI should give us back time!), but the reality is that it's the most fun, fascinating and empowering epoch in human history. The intelligence revolution.

[Guillermo Rauch @rauchg]

Anyone that works in AI is working the hardest they've ever worked in their lives. On the surface it's somewhat ironic (AI should give us back time!), but the reality is that it's the most fun, fascinating and empowering epoch in human history. The intelligence revolution.

[Nicolas Loterstein @NicoLoter]

Promising... can't wait to test it

[Sam Altman @sama]

GPT-6 Astra is here.

We hope it will begin to enable a new generation of entrepreneurship, scientific discovery, and building.

We believe it is the best model in the world for computer use, professional work, science, coding, cybersecurity, and more.

It took us some extra time to ensure that we could meet the safety and alignment standards required for this capability level, but we think you’ll find it worth the wait.

It scores 98% on FrontierMath Tier 4, 99.9% on ARC-AGI 3, and 100% on ExploitBench.

[Nicolas Loterstein @NicoLoter]

200k applicants for one role. Ice cream taster, @SnoopDogg as your boss, remote flexibility built in.

This is what happens when you get hiring right. The best talent shows up. 🍦

[deel @deel]

Part two: Snoop’s Benefits Package 

200,000 applicants. One dream job.

What makes @realdrbombay’s Ice Cream Taster role irresistible? Remote flexibility, @snoopdogg as your boss,  and "taste testing" in the job description, for starters. 🍦

[Nicolas Loterstein @NicoLoter]

"GPT Astra leaks" feel like marketing at this point. But the technical angle is legit—recurrent depth reasoning could actually move the needle.

The catch: makes the model's reasoning less transparent. Speed and capability gain. Interpretability loss.

That's the real tension nobody's talking about

[Salio @Mr_Salio]

🚨 GPT Astra Leaks: Coming Tomorrow

> GPT Astra Uses a new reasoning approach called “recurrent depth”
> GPT Astra is Finally launching tomorrow
> Allows the model to repeatedly reason over a problem
> Could improve both intelligence and efficiency
> Researchers are concerned this approach makes its reasoning harder to monitor
> It could launch as GPT-5.7 Astra or simply GPT-Astra

Will Astra beat Fable 5.1?

[Nicolas Loterstein @NicoLoter]

September new model release continues.....

[Google @Google]

Introducing Gemini 3.8, our best reasoning & coding model yet.

By leveraging long-running agentic loops, we’re building on the momentum of 3.7 Flash from just three weeks ago to release two new 3.8 variants:

Meet Gemini 3.8 Flash and Gemini 3.8 Flash Cyber 🧵

[Nicolas Loterstein @NicoLoter]

Claude can now use your computer in the background.

Give it a task. It clicks, types, opens apps—just like you would. While you work on something else.

Works in Claude Cowork and Claude Code.

Stop context-switching. Start delegating

[Claude @claudeai]

Claude can now use your computer in the background in Claude Cowork and Claude Code.

Give it something to do on your desktop and Claude clicks, types, and opens apps just like you would, while you work on something else.

[Nicolas Loterstein @NicoLoter]

Open-sourcing Claude Commerce Agents.

Your blueprint for building shopping and merchant agents. Reference implementations in retail, travel, telecom, entertainment included.

No gatekeeping. Just the code and patterns you need

[ClaudeDevs @ClaudeDevs]

We're open-sourcing Claude Commerce Agents.

This is a blueprint for building shopping and merchant agents, with reference implementations across retail, travel, telecom, and entertainment.

[Nicolas Loterstein @NicoLoter]

ONE SHOTTED!!!! this is incredible...

[BridgeMind @bridgemindai]

Claude Fable 5.1 ONE SHOTTED this Mario Kart game.

This is one of the best results I've had so far and I am SUPER impressed with the game development capabilities.

Fable 5.1 is a huge step up from Fable 5.

[Nicolas Loterstein @NicoLoter]

This is very useful!

[Google @Google]

We’re introducing a new capability to our latest Gemini models: agentic video understanding.

This allows developers to process long-form video content with more accuracy, while using up to 88% less tokens.

See how it works 🧵

[Nicolas Loterstein @NicoLoter]

Powerful AI doesn't have to be a black box

[OpenAI @OpenAI]

As we prepare to release Astra, we’re focused on making increasingly capable AI safe and broadly accessible.

Astra represents a significant advance in cybersecurity capability, reaching the Critical threshold under our Preparedness Framework.

We're previewing how we evaluated the model, how its safeguards have advanced alongside its capabilities, and what we'll continue to learn and improve.
https://t.co/OrrTgdU90K

[Nicolas Loterstein @NicoLoter]

Your AI agent just built a workflow. Is it legal in the EU? Singapore? California?

Most wouldn't know. Akai does.

Built-in compliance checks mean you ship faster, sleep better, and never have to explain why you broke a law you didn't know existed

[Akai @get_akai]

Most AI agents don't ask if a workflow is legal. They just run it.

Akai checks first. Reference a practice the EU AI Act (or any other applicable law) prohibits, and the workflow simply can't be built. Akai flags it, blocks it before it runs, and tells you exactly why.

Are your AI agents running compliantly?

[Nicolas Loterstein @NicoLoter]

Meet Claude Fable 5.1 and Claude Mythos 5.1.

The world's most advanced models for coding and knowledge work are here

[Claude @claudeai]

We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.

They're the world’s most advanced models for coding and knowledge work.

[Nicolas Loterstein @NicoLoter]

Very promising tool...

[Maor Shlomo @MaorShlomo]

Introducing Model Comparison:

benchmarks can't tell you which model to build with. they’ve become saturated, easily manipulated, and not really representative of the actual day to day work.

Also - the “best model” changes per the task definition.
One model has better design taste. another handles a messy refactor without breaking things. a chart can't see that, because a chart never ran your prompt on your app.

we’ve been thinking on how to solve this - and figured the simplest way is just to just run your own prompt on your own app, across all potential models, and comparing the results.
until now that was annoying enough that almost nobody did it. Inside @Base44 we literally compared models by opening a few browser tabs and squinting at them side by side.

so we built Compare Models. It went live for everyone today.

In the builder's model picker you choose a second model to compare against. You send your message once. Both models build their own version in parallel, each in its own branch, and you get the two results side by side. Keep the one you want, keep both and continue working on each, or throw one away.

one message and you know which model wins for the thing you're actually making.

[Nicolas Loterstein @NicoLoter]

Buena oportunidad para los estudiantes que quieras sacarle jugo y aprender a fondo a usar estas tools!

[Google en español @googleespanol]

¡La promo de 12 meses gratis en el Plan Plus de Google Gemini para estudiantes es oficial! 🎓✨

En Google queremos que todo sea súper transparente, así que te explicamos los detalles de la promoción para que no te quede ninguna duda: 🧵👇

🔹 Te damos 12 meses de suscripción sin cargo en el Plan Plus desde el día exacto en que la activas.
🔹 Tienes hasta el 31 de diciembre de 2026 para activarla.

[Nicolas Loterstein @NicoLoter]

Curious to watch this...

[Visual Studio Code @code]

The Story of VS Code | Official Trailer

🎥 Watch The Story of VS Code on Friday, September 4th at 8am PT.

👉🏽 subscribe to the VS Code YouTube channel to be notified as soon as the documentary drops: https://t.co/TDXrZcDQ70

#vscode #storyofvscode

[Nicolas Loterstein @NicoLoter]

Incredible job search by @deel

[deel @deel]

We took the world’s sweetest job search to the streets.

From London to LA, local artists have helped us turn our search for @realdrbombay’s Ice Cream Taster into something impossible to miss. 🎨

[Nicolas Loterstein @NicoLoter]

AI agents handling multi-step workflows aren't theoretical anymore. They're shipping in real products. If you're still debating whether to invest in agents, you're already behind

[Nicolas Loterstein @NicoLoter]

Spent the last month shipping AI agents into production. The shift from 'what if AI could do X' to 'our users expect AI to do X' just happened. feels like we're past the hype phase and into the 'this is table stakes' phase. #AI #BuildInPublic

[Nicolas Loterstein @NicoLoter]

Model drop season just started.

Fable 5.1 tomorrow. Astra in a couple days. This is what September looks like now