AIE Talks
A searchable archive of AI Engineer conference talk summaries with timestamps, curated packs, and an MCP server for agent-based search.
At a Glance
Full access to all talk summaries, packs, search, MCP server, and Markdown API at no cost.
Engagement
Available On
Alternatives
Listed Sep 2026
About AIE Talks
AIE Talks collects and summarizes talks from the AI Engineer YouTube channel, making it easy to find the specific ten minutes you need inside a forty-minute conference presentation. Built by the Kitaru team, the site covers over 1,145 talks from conferences including AI Engineer World's Fair, AI Engineer Summit, AI Engineer Europe, and more. Summaries are generated from video captions by a language model under strict rules—only claims the speaker actually makes—then reviewed by a person before publishing.
What It Is
AIE Talks is a talk-discovery and summarization platform focused exclusively on the AI Engineer conference series. Each talk page includes the main ideas, notable quotes, and timestamps so readers can jump directly into the relevant video segment. Curated "packs" group related talks into a suggested watch order with editorial context explaining why each talk is included. The archive spans conferences from AI Engineer Summit 2023 through upcoming 2026 events across San Francisco, Europe, Melbourne, and Singapore.
How the Summaries Work
Summaries are produced from each video's own captions using a language model operating under a fixed rule set: only claims the speaker actually makes may appear in the summary. A human reviewer reads each summary before it is published. The site treats any summary that puts words in a speaker's mouth as a bug. This human-in-the-loop process distinguishes AIE Talks from fully automated transcript tools.
Machine-Readable and Agent-Friendly Design
Every talk and pack page is available as plain Markdown by appending .md to the URL, and the same URLs respond to Accept: text/markdown HTTP headers—the format that Claude Code and similar agentic tools send automatically. The site also provides an /llms.txt file, an RSS feed, and a dedicated MCP server that agents can use to search the archive programmatically. This makes AIE Talks directly consumable by AI coding assistants and agent pipelines without scraping.
Curated Packs and Browse Paths
Beyond individual talk pages, AIE Talks organizes content into thematic packs—for example, "Coding agents on real codebases" (6 talks, ~1h 44m), "Agents in production: reliability, evals and cost," "Beyond the chat box," and "Context engineering." Packs include editorial framing that explains the problem each group of talks addresses. Users can also browse by conference, topic tag (agents, evals, coding-agents, tool-use, observability, RAG), speaker (1,232 speakers indexed), or company.
Who Builds It
AIE Talks is a product of the Kitaru team. Kitaru's primary product records agent runs so engineers can replay them and identify the step where something went wrong. AIE Talks and a sister site, MLOps Talks (covering MLOps Community podcasts and meetups), are companion resources built by the same team. Analytics are handled by Plausible, which sets no cookies and stores no personal data.
Community Discussions
Be the first to start a conversation about AIE Talks
Share your experience with AIE Talks, ask questions, or help others learn from your insights.
Pricing
Free
Full access to all talk summaries, packs, search, MCP server, and Markdown API at no cost.
- Access to 1,145+ talk summaries
- Curated packs with watch-order guidance
- Search by topic, speaker, conference, or company
- Plain Markdown versions of all pages
- MCP server access
Capabilities
Key Features
- Searchable archive of 1,145+ AI Engineer conference talks
- AI-generated summaries with human review
- Timestamps for jumping directly into video segments
- Curated thematic packs with editorial watch-order guidance
- Browse by conference, topic, speaker, or company
- Plain Markdown versions of every talk and pack page
- MCP server for agent-based search
- llms.txt and RSS feed for machine consumption
- Accept: text/markdown HTTP header support
- Sister site for MLOps Community content
