vidtheque.

following AI Engineer · 310 talks watched

the knowledge of the builders, on tap

Builders talk.
Your agent listens.

Behind this page is every talk AI Engineer published in 2026, more conference than anyone has time for. Ask any AI engineering question, your agent answers from what was spoken and what was shown, down to the second.

ready

Open the demo

the light tablenothing lifted yet

the light tableseen — owl:FunctionalProperty · line 11 of 34 · conf 1.00src Sir59K8ZDPU · tc 00:18:56

Matched keyframe at 18:56 of Why Agentic Systems Need Ontologies — Frank Coyle

heard — spoken at 00:19:01

So you have these functional properties, disjoint properties … the errors it can catch, look over in the right-hand column.

Frank CoyleWhy Agentic Systems Need Ontologies

youtu.be/Sir59K8ZDPU?t=1136ocr · exact · 15 frames · 35 cues · 3 videos

the videos belong to the people who made them

four stills, off the wall

You don't have time to watch it all. Your agent does.

Every still below is a moment your agent can hand back mid-task: the sentence a builder actually said, the text the machine read off the screen behind them, and a link that lands on the second. When a moment has nothing to give, it says so — it never quietly narrows the answer.

heard — every sentence spoken, aligned to the word.

seen — every line that crossed the screen, kept with the box it was read from.

the frame — kept as evidence, even when nothing on it is readable.

still nº3285 · ord 29 · 09:26 of 20:35

Keyframe at 09:26 of From Signal to PR: Anatomy of a Self-Improving Agent — Jason Lopatecki

Jason LopateckiFrom Signal to PR: Anatomy of a Self-Improving Agent · Arize

Can I start with like all this information on the issue and guide it the rest of the way is kind of where we are right now.

seen

The pull request Guard todo_update against in-progress items + if item.status == "in_progress": + return # already running don't retry

youtu.be/9HbzAWnKbo4?t=566a pull request rendered on a slide

still nº0041 · ord 40 · 05:27 of 18:25

Keyframe at 05:27 of Building pi in a World of Slop — Mario Zechner

Mario ZechnerBuilding pi in a World of Slop

A form two thesis is, we are in the fuck around and find out phase of coding agents, and their current form is not their final form, right?

seen

We are in the fuck around and find out phase of (coding) agents

youtu.be/RjfbvDXpFls?t=327the slide agrees with the speaker

still nº1944 · ord 59 · 11:25 of 21:32

Keyframe at 11:25 of fighting slop with slop — Vaibhav Gupta

Vaibhav Guptafighting slop with slop · Boundary

And I think if we go start thinking about it from this way, it's not so much about reading all the code, but it's more so about as a human understanding the system that you're working with.

seen

<repl>.ts:7:5 - error TS2322: Type 'number' is not assignable to type 'string'. let x: string = 1;

youtu.be/AMiyLItEtLA?t=685a REPL, mid-demo, four commands deep

still nº0564 · ord 73 · 24:52 of 43:20

Keyframe at 24:52 of Local Models: Trust, Control, Optimization — Carter Abdallah

Carter AbdallahLocal Models: Trust, Control, Optimization · NVIDIA

So there's a reason why chatbot isn't good at self-driving, because it's not trained on that, it's not deployed into that context.

youtu.be/FWMJQDH3iK0?t=1492no on-screen text at all — the frame is a person, the evidence is the sentence

talks: 310
hours watched: 107h 13m
moments kept: 19,524
lines read off the screen: 336,843

69,080 sentences spoken · 1,026,030 words · 12,855 frames read

the wall, running · 70 keyframes

Follow the builders. Get the sentence, the slide, and the second.

Every frame below comes from a talk taken in whole. Point it at a conference, a channel, a creator, and their knowledge compounds into yours.

one channel followed so far — AI Engineerpublished 2026-01-06 → 2026-08-11 · click a frame to open the talk at that second

+ follow a channeltoday: point it at a channel. next: it keeps watching.

the booth log · recorded 2026-08-09

Your agent watched it.

An agent, on the real protocol, answering a question no single talk answers — with citations, five speakers deep.

21 callsclaude sonnet

What do speakers disagree about when it comes to LLM as a judge?

call search {"q":"LLM as a judge","content_type":"transcript","max_per_video":1,"limit":6}

← 6 moments · 6 talks · leg: transcript · relevance-first

They agree it is the default and disagree about whether it deserves to be. Two positions, and a third group that ships it anyway:

“And LLM as a judge doesn't really work either because LLMs don't have good taste in writing.”

Nick Heiner · Surge AI · When Will The Benchmaxxing Plague End? · 16:01 · youtu.be/-npY6XjM8CQ?t=961

“Agent as a judge is about adaptive dynamic analysis. LLM as a judge just gives you a fixed rubric with these fixed scores.”

Aparna Dhinakaran · Arize AI · The Future of Evals · 4:07 · youtu.be/q2JrUKBMf0w?t=247

call search {"q":"LLM as a judge in production","content_type":"transcript","max_per_video":1,"limit":6}

← 6 moments · 6 talks · leg: transcript

And three teams who run it in production regardless — video at b_PmGocP4rc (Character.ai), agent evals at 31GUkCBD-Uc (Uber), clinical at O72p-rBb2bA (SonderMind). The disagreement is not “does it work”; it is what it is allowed to be the judge of.

yours, on your box

Run your own. Point your agent at it.

One compose file, one SQLite file, no build step, no runtime network dependency. It runs the CPU half on a Pi.

1 · run the corpus

$ docker compose up -d

clone, copy .env.example to .env, and it is up.

2 · hand it to your agent

$ claude mcp add --transport http vidtheque https://vidtheque.dev/mcp

one line for Claude, Codex, or anything that speaks the protocol.

macOS · Linux · MIT · runs the CPU half on a Piearly development · no releases yet · schemas can still change

Builders talk.Your agent listens.

You don't have time to watch it all. Your agent does.

Follow the builders. Get the sentence, the slide, and the second.

Your agent watched it.

Run your own. Point your agent at it.

Builders talk.
Your agent listens.