Muse Wonder might beat Astra on benchmarks, but it's nowhere close to Astra on taste. Same prompt: - Wonder (xhigh) = $2.42 in 20 min - Astra (low) = $2.19 in 7 min Benchmarks don't tell you everything 😬

For everyone asking how I made this: xhigh effort, 2 prompts. Opus 5.5 one-shot the app redesign and video. I had some nitpicks so I asked for another iteration. That's all. I gave it no direction as I wanted to see what it'd come up with on its own. It used Blender. No skill or MCP, it just figured out the scripting interface. It modeled everything itself except the iPhone, which it downloaded from Apple's website. All other video editing was done in code. --- Prompt 1 --- hey, i want you to get the entire app working perfectly on iphone duo (the upcoming foldable). i have no idea how it should work really but it should make good use of both screens and feel fluid and natural in every position the phone can be in. i installed the sdk but idk how any of it works. if you need me to approve/do things let me know but otherwise i want to see if you can just do this whole thing on your own and make it look awesome. go do some research and think about how it should all work and write up a proposal, then implement it, and shoot a really cool mind blowing demo that looks like an apple ad, like the device in various realistic scenes modeled in 3d with beautiful fluid transitions showcasing the app transitioning through various the device orientations with real footage from the sim. i want to ensure the app feels delightful and makes great use of duo, so look for opportunities to really do something cool or unique with our 3d layout on the duo form factor --- Prompt 2, after seeing the first draft --- not bad! some notes: - 3d model of the phone has some horrible jaggies around the edges - some of the transitions are ugly jumps, not smooth; clean those up - pacing is weird, takes too long to get to anything interesting like the phone actually opening and doing cool transitions in the app - sim footage is dropping frames - need better music pull out all the stops actually make this look like an apple ad

I'm stunned. I told Opus 5.5 to update my app for iPhone Duo, however it wants, and send me a cool demo It ran for 14 hours, redesigned the app to look great in every mode, and made this completely over-the-top video …for 9% quota

Forget threejs demos, Opus 5.5 is *incredible* at building apps. This is Letters Abroad, a language-learning app I'm working on. You exchange letters with an AI pen pal, and learn a language through context (plus some brain science magic). (also built by Opus, of course) I've been tinkering with this the last week. I built a prototype with Astra, but wasn't happy with the design. Tried Fable, still wasn't good enough. Then Opus 5.5 dropped. In a few hours, I've made more progress on the visuals than in a week of burning tokens. This model is insane. It reads my mind and nails it every time. All the gorgeous 3D models and fluid transitions are thanks to Opus. Crazy idea, prompt, repeat. No more quota anxiety, just non-stop flow state. I'll be writing about my prompts/skills/learnings from building this. Lots to share; this generation of models is absolutely game-changing for design!

I had Astra analyze all my Codex threads and learned a lot about quota. Here's how to get the most Astra usage. (applicable to Fable too) I've spent ~$5000 on Astra since release*: - 70% is JUST cached input reads - 16% is uncached input - 13% is output - <1% is cache writes Till now, my Codex threads kept getting longer and longer, but Astra broke the trend (see chart). Long context is eating most of your quota. The biggest levers: - Smaller threads. New thread for each little task. Don't let context build up - Use smaller models that delegate coding tasks to Astra subagents. Don't let Astra see big outer thread - Avoid continuous polling. Have Astra set timers and wake back up for long-running tasks - Trim tokens wherever you can: skills/MCPs, AGENTS.md, custom instructions, etc - 3rd party tools like RTK that return more compact outputs might help (haven't tested this yet) Investigating that 70% input cost in more detail: - 40% is encrypted content (reasoning or other OpenAI stuff) - 20% is file reads - 10% is tool call arguments, schemas, etc - 7% is skills (catalog, loading instructions, output) - 5% is base instructions (system prompt, custom instructions, AGENTS.md) - 5% is browser/computer use input - 3% is command/tool output - The rest is minor or not actionable: compaction handoffs, assistant prose, user messages What you can do: - Use the lowest reasoning you can for your task, to cut CoT tokens - Make your repo as agent-friendly as possible. Small files/modules with clear folder structure, so Astra isn't grepping and reading large chunks. Try asking Astra to delegate file search tasks to subagents like Luna - Don't dump in tons of skills, instructions, AGENTS.md stuff; be intentional about it - Delegate browser/computer use to smaller models unless you need max performance And, probably goes without saying: - Don't use /goal with Astra. The few times I did absolutely obliterated my quota - /fast mode basically only if Tibo says he's resetting in an hour and you have 100% quota to burn Let me know if you have any observations to share! *I didn't literally drop $5k, this is mapping token to equivalent API prices

These alleged GPT-6 Sol leaks look even better than Astra. We know they have a much stronger internal model. Maybe they’re distilling it? Astra is 2.5x the price of Sol. This quality at Sol prices would be completely insane. Hope we see it this week…

After trying Luna xhigh to use Astra agents for improving the game, and failing, even after passing reference images, I asked @anshuc and he helped me with an improved prompt. The result is interesting, and it took 1% of my usage. Prompt in the first comment

dude GPT-6 Astra is some kind of turbo-AGI machine god for 3D games. It one-shot this in 45 minutes for hardly a couple % of my quota. I figured out how to get great graphics out of it. The trick is image gen. I'll share the process below.

Mark my words, Astra is foreshadowing what the OpenAI device will be like. This is OpenAI’s vision for AGI. The form-factor for AGI will be intelligence that kinda just sits on your shoulder, sees what you see, anticipates your needs, and gets things done. OpenAI almost has all the components to build this: - Astra: a model that’s fast, smart, capable of multitasking - ChatGPT Work: cloud infra for horizontally scaling agents with their own computers - Computer Use: a way to bridge the gap between LLMs and all the messy software humans use - GPT Live: a bidi voice architecture that can run continuously, staying quiet or jumping in as appropriate - Hardware division building some mystery wearable These are the Exodia pieces, and when they come together, it’ll change our relationship with technology. Most computers will only exist for agents. Most work will happen by talking to your OpenAI pendant or whatever, which will orchestrate dozens of agents across virtual computers you never see or think about. Many challenges remain (alignment, embodiment, memory, battery life, …), but this seems as tractable—even imminent—as ever. Scaling just keeps working. The internal branches are converging for OpenAI.

Fable 5.1 is an absolute beast for game dev. I took my Opus 5 space game and told Fable to "just make it look better". It overhauled everything with a TON of detail, AND made it 2x faster 🤯 Code: Demo: Every pixel you see is made by AI. Fable modeled everything in Blender, generated textures with code, and in a few cases ran image models via CLI. In mere months we'll be vibe-coding AAA graphics at this rate. Just insane. Of course, Fable recorded and edited this comparison video too (with a little help from Opus at the end, because of the Fable outage)

Thank you all for the words of encouragement and excitement about this article! I’m glad it’s helping people think differently about how to design with LLMs. I’m working on follow-ups that go deeper on these ideas, so keep an eye out for those both here and on Substack. This has been a true labor of love: many late nights and early mornings editing drafts from planes, trains, and coffee shops across Tokyo, Manila, SF, NY, and Montreal. Huge thanks to @lennysan and his incredible team. They have been a dream to work with.

I'd always thought AI was terrible at design, but after reading today's 🤯 post by @anshuc, I realized I was just doing it wrong. "AI models are capable of amazing creativity, but that creativity gets stifled. LLMs are trained to be next-token predictors: they look at a sequence of text and predict what typically comes next. Great design is exactly the opposite of this. Great design bends the rules and delights users with memorable, unexpected choices." @anshuc led design and engineering teams at Apple for 12 years. In his words: "Most people only see 1% of AI's creative potential. I want to show you how to tap into the other 99%." His 8 techniques for breaking out of the 1%: 1. Use seed strings to inject variety 2. Be much more ambitious with your prompts 3. Create positive feedback loops with subagents 4. Use image generation to enrich designs 5. Use video generation 6. Cut out elements that don’t add value 7. Remove AI tells 8. Rewrite copy by hand Read the post here: P.S. This design was made by AI 👇