Why does frontier AI still forget everything after a short break? You can have a solid 2-hour working session, step away for a few hours (or until tomorrow), and even with a good recap it only remembers a fraction of what actually mattered. I’ve tried every workaround — operating instructions, session recaps, screenshots, external memory files — and they all still feel like manual babysitting. Is this just a hard technical problem… or are the big labs keeping long-term memory weaker on purpose because it burns more tokens and keeps you engaged longer? Full breakdown + my honest take: trainingrun.ai/day-009.html @karpathy @AnthropicAI @OpenAI @xai What actual memory techniques or tools are you using that work reliably when you come back hours or days later? Drop your best ones below — I read every reply. — David Solomon TrainingRun.AI/news

🚨 The Pentagon just labeled Anthropic (Claude’s makers) a “supply-chain risk.” Why? Anthropic said NO to letting their AI be used for: Fully autonomous killer weapons (no human in the loop) Mass surveillance on Americans Pentagon basically said: “That stance hurts national security.” This is the first time a major U.S. AI company has been hit with this label — the kind we usually reserve for foreign threats. Full plain-English breakdown + why this matters to every family: trainingrun.ai/day-008.html What do you think? Should companies be forced to say yes to military use, or is “no” the right call? I read every reply. This isn’t just tech — it’s our future. #AI #Ethics #Anthropic #Pentagon @AnthropicAI @Dario_Amodei @elonmusk @karpathy

🚨 New trick just dropped that makes Grok 4.20 (and every other coding agent) way more trustworthy. Instead of guessing about code, agents now have to show their work like a careful engineer — no dangerous execution needed. Teach your kids how to be an engineer! Accuracy jumps to 93% on real patches. Safer, cheaper, and humans stay in control. Full plain-English breakdown + why this helps all of us decide how AI should shape our future: trainingrun.ai/news @Meta @xai @grok @elonmusk @karpathy #AI #Agents #Grok420 #CodingAgents