Day 16/ You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache.

1/ Hermes can now launch Cua Spaces. When a task needs a computer, Hermes creates a Space on your Mac, Windows PC or Linux box, does the work there, and deletes it when it's done. You and your team can jump into the same Space while it works.

I've been using my transcription app for a few weeks now, at work and outside of it, and I think it's time to introduce it properly. It's called Takku, and this is the site: takku.app It has replaced Granola for me, and I'm not going back. It's lighter, it's faster, and I don't run into the limits of Granola's free plan. It also costs next to nothing: with GPT-6 Luna, Haiku 5.5 or DeepSeek, an hour-long meeting costs cents. I optimized it for the Mac and didn't add any useless features. I want it to do one thing and do it well. The Unix principle. The latest addition is Hooks: when a meeting is ready, Takku runs a script of yours. That's how it plugs into the rest of your system, for example by publishing the meeting to Notion. A few colleagues and friends are already using it, and the feedback has been good so far.

in other news; i’m ditching flutter. going for rust core then native mobile/desktop apps per platform for overalls.dev

Moved to tsrs over the weekend. Drop-in replacement. Around 5x faster for OC. Kind of amazing that it took Microsoft a team and two years to rewrite the TS compiler in go, and now you can just let an agent rip and get sth significantly better within weeks.

🚀 Introducing DeepInfer, our in-house inference framework, now fully open source! Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable. 🔗 github.com/deepseek-ai/DeepI… 1/4

i heard you guys want a herdr app...

MiniMax M3.1 Flash: Space Performance + Bunny Speed 🚀🐰 Yes, Space Bunny - which has topped global leaderboards with 200T+ tokens - was an earlier version of M3.1 Flash. 💰Redefining the performance economics of agentic AI. M3.1 Flash achieves frontier-level intelligence across key tasks such as Cowork (Automation-Bench: 62%; ALE: 30%), Coding (DeepSWE: 74%) and Visual Reasoning (BabyVision: 91%), at 1/3 the cost of M3 and 3x the speed. API pricing (flat rate): $0.15 / $0.60 per 1M input/output tokens, with cache reads at $0.015 / 1M tokens. Model weights will arrive on Oct 23. API: platform.minimax.io M Plan: minimax.io/m-plan MiniMax Code for desktop: agent.minimax.io/download

DeepSeek V4.1 Flash 1017 just matched Opus 5.5 on long horizon browser use. 19x cheaper 👀 > 75/100, same as Opus 5.5, almost Astra level > $0.22/task vs $2.35/task > Sol it still SOTA, but it's close The new pareto frontier 🔥

Prompt: A highly realistic mirror selfie of a stylish young woman inside a modern elevator with warm metallic bronze walls. She has a short, wavy brunette bob with natural highlights, soft facial features, and a playful expression as she closes one eye and makes a cute kissy face toward her phone. She wears a fitted royal-blue cropped T-shirt, light-wash high-waisted jeans, and a charcoal-gray sweater draped casually over her shoulders. A delicate gold necklace, dark burgundy nail polish, subtle tattoos on her arms, and a sleek black leather handbag complete her outfit. She holds a light-blue smartphone in one hand while posing naturally in front of the elevator mirror. Warm ambient lighting, realistic skin texture, natural facial details, authentic body proportions, crisp fabric textures, accurate reflections, elegant casual fashion aesthetic, shot on a high-end smartphone camera, ultra-sharp focus, HDR, 8K quality, photorealistic, candid -style photography, no artificial beauty filters, no CGI, no text, no watermark.