Trails

Start here

Four short paths through 946 talks. Pick the problem you have — not the hype cycle.

Path 1

Evals first

If agents are non-deterministic, measurement is the product. Start here before you scale prompts or skills.

1

The stake in the ground: don’t ship agent skills without a way to measure them.

Don't Ship Skills Without Evals — Philipp Schmid, Google DeepMind
2

Hands-on: what evals look like when the unit of work is an agent, not a completion.

Ship Real Agents: Hands-On Evals for Agentic Applications — Laurie Voss, Arize
3

The category error — why unit-test instincts fail for stochastic systems.

Evals Are Not Unit Tests — Ido Pesok, Vercel v0
4

Field notes from running evals in production: what actually breaks.

Five hard earned lessons about Evals — Ankur Goyal, Braintrust
5

Where the practice is headed once measurement is table stakes.

The Future of Evals - Ankur Goyal, Braintrust

Path 2

Harnesses, skills, and taste — how teams actually ship with Cursor, Claude Code, and friends.

Path 3

From protocol origins to production MCP — what survives contact with real users.

1

Protocol intent from the people shaping MCP’s next chapter.

The Future of MCP — David Soria Parra, Anthropic
2

Remote MCP in the wild: what shipping taught Anthropic.

Remote MCPs: What we learned from shipping — John Welsh, Anthropic
3

The skeptic’s cut — what’s still broken before you bet the architecture.

MCP Is Not Good Yet — David Cramer, Sentry
4

Skills + MCP together: closing the context gap in practice.

Combine Skills and MCP to Close the Context Gap — Pedro Rodrigues, Supabase
5

Security is not optional: insecure MCP dies in production.

Your Insecure MCP Server Won't Survive Production — Tun Shwe, Lenses

Path 4

For builders who care about judgment, UX, and getting something into someone else's hands.