JonathanCochran

The singularity won’t feel like a takeover. It’ll feel like waking up healthier, smarter, and more human than ever.

GPT-5.4 mini is available today in ChatGPT, Codex, and the API. Optimized for coding, computer use, multimodal understanding, and subagents. And it’s 2x faster than GPT-5 mini. openai.com/index/introduc…

@alexwg @marshallbrain’s Manna feels less like sci-fi every year. When task guidance and behavioral scoring move into headsets, management becomes software. marshallbrain.com/manna1

@flowersslop x.com/getalby/status…

has someone tried to give clawdbot its own clawdbot?

@chatgpt21 This one’s been useful for me as well, for similar reasons: spicylemonade.github.io/AI-2027-tracke…

I like this a lot. I’ve seen a few takes on the AI-2027 framing, but the milestone matrix and readiness weighting make this feel more grounded than most. The minimalist design fits the source material well, too. I’d be in favor of releasing it, tools like this help turn speculation into something you can actually reason about.

I built this 2027 tracker demo in GPT 5.3 to bridge the gap between speculative hype and measurable progress. The dashboard I made utilizes a tiered Milestone Matrix to track 14 mission-critical KPIs from the OG paper. I also wanted it to weight "Readiness %". The design is very intently minimalist to match the original feel of the AI2027 forecast. But if you look closely there are a few creative design choices 5.3 thought of to show case the differences. Let me know if I should release it 😊

@flowersslop Agreed, games are underrated as agent training grounds. Partial information + competition tends to surface very different behaviors quickly, even with similar models.

AI vs AI bomberman training sandbox, one attempt, and it works really well. both agents quickly learn the game and try to outplay each other. codex 5.3 xhigh using pygame and q learning