@waveshare00 Awesome Device

Now I have seen it all, Effective November 12th, 2026, abusive behavior towards Claude models will be a violation of Anthropic's Usage Policy. I might have to encourage people to say please and thank you again 🤣

embers.assembll.xyz/ now in free beta to play - Just give me feedback! Embers Season 3 is live 🔥 A browser survival game where an AI overseer, Ember, judges how you play. Crimes only count if someone sees them. Outlaws get hunted by AI bounty hunters and helicopters. Richest team wins the wipe. Build, trade, Raid and Survive

OK new update to embers.assembll.xyz/ I wanted to focus more on a actually playable game rather than a fancy demo, so I leverage @Polyfork3d for new player models, assets and kits which helped improve player dynamics, like rolling, jumping etc. Also add @OpenAI Luna for NPC's in the game. Now there is quests, events and factions.

OK here’s my key takeaways from OpenAI DevDay full recap: openai.com/index/devday-2026… Dots Grok Bot, Muse, and now dots look similar. they’re not. OpenAI’s version is the more cohesive one across personal + business, and single-player + multiplayer. each dot runs on GPT-6 Astra, has its own cloud computer, and connects to 4,000+ apps. pretty excited to see where it goes. openai.com/index/introducing… help.openai.com/en/articles/… Sol 6.1 biggest thing? cheaper cached tokens. $0.10 / 1M cached input. that actually moves the cost of long-horizon work. also weird that 6.1 dropped a few days after Sol 6. feels like 6 was premature and 6.1 is what 6 was supposed to be. openai.com/index/introducing… Decisions api incredible how fast they shipped a structured decision layer. also a reminder that frontier labs can move into a new category very quickly. OpenAI already has multimodal models in the same stack. that’s the moat. openai.com/index/devday-2026… openai.com/index/gpt-6-astra… Ultrafast 8x speed at 6x price on Astra. Sol coming later. how many workflows actually get positive ROI on Astra at 6x? developers.openai.com/api/do… Sign-in with chatgpt + marketplace this is the big one. devs can ship agentic products without selling tokens. customers buy compute from OpenAI and use it in third-party tools. selling tokens is a bad SaaS model. this is better for the ecosystem. developers.openai.com/siwc help.openai.com/en/articles/… developers.openai.com/plugin… curious what you took away.

Made a @playrust inspired Online Multiplayer game with @claudeai Opus 5.5 and @threejs - The most impress part of this game is the game procedural graphics and dynamics with crafting, building, dynamic weather and events, NPC's and animals controlled by AI. It ran for about 2 days off one prompt and subagents and I didn't even burn the Max plan on Claude. embers.assembll.xyz/ Check it out and play, it still have some rough edges, but would love any feedback, also happy to share the repo with anyone interested.

Looks like a decent direction for CoPilot, focusing more on the product and harness rather than a model could be good, with all things Microsoft we will see how good the execution is

I think everyone needs to take a breath before jumping to conclusions that a rogue agent decided to hack a government site. The more likely occurance was the agents was asked to perform web based research by a users and was looking around sites and found an area that was unprotected.

A model shipped this month that can't write a sentence. That's the point. Give it a situation + a fixed list of answers. It points at one, with a calibrated confidence score. ~100ms, fractions of a cent.

This model is fantastic for certain use cases, super fast