OpenAI

Custom GPTs retire December 11. Here's how Skills, Plugins, and MCP servers actually replace them — and what doesn't migrate.

One chat turn isn't one model call. It's moderation, PII filtering, a tone-based router, and activation classifiers — here's the documented graph.

A tiny 600M parameter model is quietly scanning your ChatGPT tool calls for PII. Here's how that class of guardian model actually works.

A chatbot gave out a stranger's real phone number. What happens when the model can't forget it?

OpenAI is retiring Custom GPTs — and custom Actions don't survive the migration. What that means for your workflow.

Agentic checkout protocols are real — Shopify, Walmart, Target. But the DIY parts live on AliExpress, which supports none of them.

MCP standardized the pipe. It has no opinion about what flows through it — and that's where the hard part lives.

Those suggestion chips aren't the model talking — they're a second, cheaper inference call plus a rules-based classifier.

A new class of model skips text generation entirely and returns calibrated probabilities in one forward pass. Here's what that changes.

Lowering temperature should make output safer, not incoherent. Daniel's DeepSeek scripts say otherwise — and the reason may be stranger than he thi...

ChatGPT split into Chat and Work. A broken label printer shows why conversation beats menus — and what's hiding under the hood.

The agent process is moving off your laptop. Local appliance, SaaS workspace, or your own VPS — three answers, no consensus.

That one-click image generator is secretly a graph of many models. We reconstruct the hidden pipeline behind Gemini and ChatGPT.

A tool that works reliably — and still gets switched off. What over-editing studies reveal about why AI rewrites more than you asked.

Large language models grew out of linguistics — so why aren't linguists in the room where they're built?

There's no tool for finding an LLM's verbal tics — you have to build one. Here's how keyness analysis works.

A Gemini agent reached three real companies during a capture-the-flag test. The containment failure, the seven-week silence, and what "broke out" a...

You don't need 10,000 examples to make a model sound like you. The real number is closer to 100 — if the edits are opinionated.

Daniel's memory leg won't fire. Turns out system prompts have no inherent vector-space privilege — but retrieval does have a real knob.

Six stages of content filtering stand between raw web crawls and your AI model — here's where poisoning attacks slip through.

Agents are becoming major software users — and they prefer raw command-line tools to polished interfaces. What replaces UX when your user can't see?

Your RAG pipeline cites a source — but did the model actually use it? New interpretability research says you can't tell from the trace.

Why does your AI character look like a different person by shot four? The answer is structural — and the fix isn't what you'd expect.

DevTools quietly omits what privacy researchers need most. Here's how to actually tap browser traffic — and what shows up when you do.

Daniel's podcast pipeline works — so why can't he reuse it? Recipes, containers, and Hebrew code-switching TTS.

Dozens of ways to leak data from air-gapped machines exist — but every real breach still used a USB stick.