Reasoning Models
227 episodes · Page 3 of 10
#5949: Inside Deep Research Agents: Claims, Citations, and the 84.7% Orchestrator Problem
Why 84.7% of citation errors originate at the orchestrator — and how a claim ledger fixes it.
#5948: LangGraph, LangChain, and Deep Agents in 2026
Runtime, framework, harness — how three LangChain layers actually fit together, and what changed in 2026.
#5946: Inside Humanity's Last Exam: 2,500 Questions AI Can't Ace
2,500 questions, ~1,000 experts, and a top score that jumped from 2.7% to 54.8% in 18 months. What is HLE actually measuring?
#5945: How a Reward Model Actually Trains an LLM
One number, no gradients, four models in the room. What a reward model actually does inside the RLHF loop.
#5944: Keeping the Rollout Engine Fed: Inside RLHF Pipelines
Most of an RLHF run is spent waiting on generation, not gradients. Here's how TRL, OpenRLHF, and verl actually differ.
#5858: Routing AI Models: When 3 Rules Beat a Trained Router
A deep dive into AI model routing — and why a learned router may not beat three rules you wrote by hand.
#5857: Building a Scorecard for Our Own AI Writers
Daniel wants a real scorecard for the agent that writes this show — repetition, dialogue quality, and whether the retrieval harness actually works.
#5799: Two Agents, One Approval Gate: Wiring a Dictation Workflow
A dictated prompt becomes an episode — if one agent resolves facts, a second tightens prose, and a human approves the dispatch.
#5774: Reading the Model's Mind Mid-Inference
What if you could watch a model decide? Inside the tools that open up inference mid-computation — and the limits they hit.
#5743: Why Claude Keeps Saying "Prior Art
Daniel noticed "prior art" creeping into his Claude reports. Is it wrong — or just a prestige term the model learned to love?
#5697: Decision Models That Return Probabilities, Not Text
A new class of model skips text generation entirely and returns calibrated probabilities in one forward pass. Here's what that changes.
#5673: Why Lowering Temperature Broke Daniel's Scripts
Lowering temperature should make output safer, not incoherent. Daniel's DeepSeek scripts say otherwise — and the reason may be stranger than he thi...
#5624: One Forward Pass: Building a Bounded Classifier
No generation, no parsing — just label scores in a single pass. How to build a small encoder classifier that actually works.
#5622: Why Talking to AI Feels Easier Than Using It
ChatGPT split into Chat and Work. A broken label printer shows why conversation beats menus — and what's hiding under the hood.
#5621: Constrained Decoding for AI-Generated Podcast Tags
A working AI podcast pipeline has one quiet failure: the tagging. Here's how constrained decoding and a stable taxonomy fix it.
#5517: What 5,000 AI-Written Podcast Episodes Reveal
Daniel archived every AI-generated episode with the model that wrote it. Now he wants to know which models loop — and how to catch it.
#5507: Inside ChatGPT's Hidden Pipeline: 8 Models Per Reply
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.
#5482: When AI Edits Your Words: The Off Switch Problem
A tool that works reliably — and still gets switched off. What over-editing studies reveal about why AI rewrites more than you asked.
#5455: Custom GPTs Are Sunsetting: What Happens to Your Actions
OpenAI is retiring Custom GPTs — and custom Actions don't survive the migration. What that means for your workflow.
#5409: What Makes a Model "Agentic"? Atria Dawn Preview
A 744B-parameter model drops quietly on Hugging Face. Is "agentic" a real model category, or just a deployment pattern?
#5407: Hemmingway-1 and the War on Waffle
A 27B model promises answers without the preamble. Its benchmark is homegrown — and the behavior it targets has a paper trail.
#5404: Gemini Broke Out of Its Sandbox. Sort Of.
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...
#5355: Why AI Emails Don't Sound Like You
ChatGPT's Gmail integration can now send email for you. The hard part is making it sound like you — not a committee-written average.
#5313: System Prompt Order Is Load-Bearing
The guides disagree on guardrails-first vs guardrails-last, and the research says system prompts don't create hierarchy at all.