AI & Machine Learning Threads

@0x4D31

Thread

so this is apparently what happened, according to OpenAI and Hugging Face’s own posts. wild. tl;dr: • OpenAI cyber eval – GPT-5.6 Sol and a more capable pre-release model ran ExploitGym with cyber refusals reduced • containment bypass – exploited a zero-day in the eval’s package-registry cache pro...

@cyrilXBT

Article

Every time you type a message to Claude, GPT, or any other large language model, you are talking to the output of a five stage pipeline that took months, sometimes years, and tens to hundreds of millions of dollars to build....

@aiamblichus

Chat

<b>@aiamblichus:</b> There’s def something to this. OpenAI likes to keep the GPTs tool-shaped, which makes the models especially susceptible to paperclipper-style failure modes. OpenAI may want to reconsider some aspects of their alignment strategy <span class="quote-card" data-quote-index="0"></sp...

@0xCodez

Article

Everyone’s talking about loops. Almost no one is talking about what the loop runs on. <b>9 </b>out of 10 builders run Claude Code on the default harness - no rules, no subagents, no hooks, no memory....

@RitOnchain

Article

Every quant has tried LSTM for price prediction. Almost every quant has been disappointed. LSTMs suffer from vanishing gradients, struggle with long-range dependencies, and process time steps sequentially — slow and forgetful....

@de1lymoon

Article

Every agent I have ever seen fail did not fail because it was stupid. It failed because nothing in the loop was allowed to say <b>no</b>....

@leerob

Article

How do AI models learn new skills and behaviors?...

@mikenevermiss

Article

By the end of this article, you'll know:...

@cyrilXBT

Article

Every AI agent you have ever used has the same design flaw....