Wednesday, September 9, 2026

Good morning, NOLA. Today is a proper builder’s buffet: Meta has launched Muse, OpenAI has a new ChatGPT Images 2.5, and a tiny GitHub skill called I-have-ADHD tackles one of coding agents’ most universal bad habits. There’s also a compelling look at AI helping run quantum-computing experiments, plus enough creative experiments to make you want to hand an image model a strange prompt over lunch.

Tools That Get Out of Your Way

Meta introduces Muse, its personal AI agent

Muse is Meta’s new personal AI agent, aimed at helping people turn everyday intent into completed tasks rather than another blank chat box. The useful question for builders is whether its personal context and task-oriented design make agents feel more like dependable assistants; the Hacker News discussion is a good early read on that.
Meta AI

ChatGPT Images 2.5 arrives

OpenAI’s ChatGPT Images 2.5 is the day’s most immediately tryable release: put an image or an idea in, then iterate toward something usable. If visuals are part of your product, pitch, event flyer, or prototype, this is the kind of upgrade worth testing against a real task today; here’s the HN conversation too.
OpenAI

A tiny skill that makes coding agents lead with the answer

I-have-ADHD is a deliberately named instruction set for stopping coding agents from burying the result under a long work log. It is a small, practical reminder that good agent UX is often prompt and workflow design, not a fancier model; see the Show HN discussion for ideas people are adapting.
GitHub

LibreOffice’s no-AI stance becomes a product signal

LibreOffice reportedly hit a download record after emphasizing that it has no AI features. This is less an anti-AI story than a useful product lesson: people also value tools that are predictable, private, and not trying to assist with every click. HN has a lively discussion on the signal users are sending.
Manual do Usuário

Astra, Put to Work

Codex helps run quantum-computing experiments

OpenAI describes using GPT-5.6 Sol to help researchers plan, run, and interpret quantum-computing experiments. The takeaway is pleasantly concrete: agents are becoming useful in real scientific workflows when they are paired with instruments, feedback, and expert review—not just asked clever questions. HN discussion.
OpenAI

The big Astra episode: capability, rollout, and practical demos

AI Daily Brief takes on why GPT-6 Astra feels consequential—and why its launch has also been confusing. Worth a listen if you want a builder-focused synthesis rather than a benchmark dump; pair it with the episode’s image-to-Lego demo for a wonderfully tangible example.
AI Daily Brief

AI has a discovery problem

This thoughtful essay argues that the hard part of the AI-tool boom is increasingly finding the right thing at the right moment. A good read for anyone building an AI product: distribution, trust, and clear moments of use may matter more than adding one more capability. HN discussion.
M. H. Acevedo

Creative Side Quests

From an idea to a buildable Lego set

Feed Astra an image or a concept and it can produce a buildable Lego set using orderable official parts. It is a delightful example of AI crossing from “nice image” into an output that can become a real object.
AI Daily Brief

A historically accurate, walkable Waterloo

A creator used the new visual capabilities to make a historically grounded 3D reconstruction of Waterloo. Whether you work in education, tourism, games, or storytelling, this is a useful prompt for imagining richer interactive explainers.
AI Daily Brief

Turn a Zillow listing into a 3D house and promo video

This experiment turns a real-estate listing into a 3D walkthrough and marketing video. The specific demo is fun, but the broader opportunity is serious: niche businesses can now prototype custom sales media without a full production pipeline.
AI Daily Brief

One photo, a local 3D house, and 60fps

Another standout visual demo: a fully navigable 3D house from one photo, running locally at 60fps. It is early, but it makes spatial product demos and lightweight virtual tours feel much closer to everyday creative tooling.
AI Daily Brief

Also

Today’s Sources