Outline — Build Your Own Product Brain

Speaker

Emma — founder/CTO of Resonant, "a product workspace for helping people write specs, PRDs, tasks, etc. and then we ship them very fast." Before founding Resonant she "was at Stripe for four years running engineering in the UK" and earlier "started my career as a PM at Google and then I ran product engineering and design at an e-commerce company." Her surname is not stated in the transcript; the slug uses a placeholder (stoneham) and should be corrected if her real surname is known.

Host: Simon Maple, Head of Developer Relations at Tessl, AI Native Dev co-host. (Introductory remarks only.)

Event: AI Native DevCon, 2026-06-01.

Abstract

Not provided by the user. [inferred]: Emma demonstrates how product managers can build a GitHub-backed "product brain" — a continuously-ingested, agent-readable knowledge graph of customer signals, Linear tickets, meeting notes, and code — and orchestrate a small set of agents on top of it to run autonomous PM workflows (triage, PRD-writing, shipping straight to coding agents) with human checkpoints for strategy and coherency.

Thesis

1. Coding has been substantially automated by agents; PM has not, and PMs are getting pulled into day-to-day work as engineers move faster.
2. The fix is to make the PM an agent orchestrator with a "product brain" — a synthesized, continuously-updated representation of the product that agents can act on.
3. The architecture has four parts (live ingestion, product frame, workflows, human input loop) plus four agents (action, input, brain, organizational), and the demo shows it routing two superficially-similar Linear tickets to two different correct outcomes.

Section TOC

Terminology glossary (definitions Emma actually gave)

• Product brain — "a reinforcement learning system that learns your product instinct" so the PM can focus on strategy and keep up with engineers. Realised as a GitHub repo with inputs/, sources/, wiki/.
• Product frame — the synthesized form of all ingested inputs "into a form that can be quickly processed by any agents that are running autonomously." (Note: Emma uses "product frame" and "product brain" partially interchangeably in places; the frame is the synthesized layer inside the brain.)
• Live ingestion — automatic processing of meetings (e.g. Granola), GitHub, email, Linear tickets so the system is "up to date" enough to run autonomous workflows.
• Human agent control plane — "How are you going to ensure that you have the right checks and balances between your agents and your PM so that they can work effectively?"
• Action agent — "It runs the workflows."
• Input / brain agent — "process those inputs and reorganize them."
• Organizational agent — keeps the wiki and team/org structure from getting "out of date" as the company reorganizes.
• Spectrum of autonomy — Emma's framing of how much a company lets agents run end-to-end vs. requiring PM checkpoints; "very much a personal taste thing that companies need to grapple with."
• Wiki (in the product brain repo) — "customer insights, features, product principles … and then technical context, which expresses our front end flows."
• Product principles — the wiki section "expressing this higher level of how do you have taste and coherency in the system."

Named frameworks / concepts introduced

1. The four-component product brain: live ingestion → product frame → workflows → human input loop.
2. The four agents: action, input, brain, organizational.
3. PM autonomy ladder (analogous to coding agents): ChatGPT workflows → skill orchestration → agents with taste/coherency/self-awareness of limits. Compared explicitly to Cursor line-by-line → Claude Code → "open Claude … if you're brave enough."
4. Two-direction PM role evolution: PMs moving down into "product builder" and/or up into "GM and strategy" roles.
5. The product brain repo layout: inputs/ (daily-processed queue, with manual/ subfolder), sources/ (raw data, not all kept), wiki/ (customer insights, features, product principles, technical context).

Open questions / not covered

• Specific evaluation metrics for whether a product brain is "working well" — not addressed.
• Exact cost/token budget of running the brain — only obliquely referenced as "cost parameters."
• How the brain handles conflicting customer signals — not addressed.
• Multi-product or platform companies — not addressed (Emma speaks from a single-product Resonant perspective).
• Onboarding flow for replacing an existing PM tool stack — Q&A touches on this but Emma says it's "an art right now."
• Concrete compliance/regulatory implementations — Emma sketches that approval flows can be enforced but says "it depends really on the underlying company."
• Pricing, licensing, or commercial details of Resonant — not addressed.
• Benchmarks vs. alternative tools (Linear AI, Notion AI, etc.) — not addressed.
• How long the product brain has been in production at design partners — not addressed.

Participants (for attribution cross-reference)

• Simon Maple — host, introduces Emma.
• Emma — main speaker (surname not stated in transcript).
• Audience questioners — none clearly named. The transcript contains "Hi Anna, can I start?" but this is likely an audience member asking as Anna, not addressing one; treat as unattributed. "I'm Pia myself" appears to be a transcription artifact (likely "I'm a PM myself"). Do not attribute questions to specific named people.
• Rob — mentioned in the demo as someone "working on" onboarding work; not a participant.