Herald Labs
Talk to the lab
Herald Labs Missions are 90-day operating sprints where Heralds build, launch, and prove a real AI-native business using the Google and Gemini stack. No simulations. No demo theatre. Ship to real users, collect evidence, and earn the mission credential.
A Herald Labs Mission is a commitment to ship. Heralds pick one of five mission categories, define a concrete business hypothesis, and use the full Google AI stack to build a product that real people use before the clock runs out.
The work is judged on what shipped, who used it, and whether AI is doing real work in production. Not on slide decks.
Each Herald owns their mission end-to-end. The lab provides the stack access, operating cadence, review gates, and peer support.
Every Herald selects one category for their 90-day mission. Categories are deliberately broad so Heralds can find a problem that matches their context, market access, and conviction.
AI-native learning tools, tutoring systems, skill verification, or access programs that help people grow capability faster than traditional paths.
Tools that help people start, run, or scale a livelihood. Think agent-assisted operations, market access, supply chain, or hiring workflows.
AI-powered services for small and growing businesses: bookkeeping, compliance, customer support, inventory, or back-office automation that saves real hours.
Products that improve how people save, borrow, transact, or understand money. Insurance, payments, lending, or financial literacy with AI doing real work.
AI systems for legal, medical, accounting, consulting, or administrative work. Domain-specific tools where accuracy, trust, and auditability matter.
Heralds are expected to use the Google AI stack as the primary engine of their mission product. The stack is not a constraint. It is the point. Build native, build deep, and let the tools do heavy lifting.
The Google and Gemini stack should be the backbone of the product, not a bolt-on. Heralds who build native on the stack learn faster and ship tighter.
The mission clock starts on day one. Each phase has a gate review where Heralds present evidence and the lab decides whether to continue, pivot, or stop.
Define the business hypothesis. Scope the thinnest end-to-end slice. Set up the Gemini-powered pipeline. Ship a private beta to a small set of target users by day 30.
Open the product to real users. Collect traffic, usage, and feedback. Iterate on the AI workflows that matter. Hit the first revenue or engagement milestone by day 60.
Harden the product. Fix failure modes. Document the human-AI labor split. Gather verifiable evidence of users, usage, and business viability.
Submit the final evidence pack. Present at the mission review. The lab scores the mission on viability, AI operations, and category impact.
Three equally weighted criteria. The scorecard is deliberately simple so Heralds know exactly what the lab is looking for and can optimize for the right things.
Is it a real business with real users, real usage, and a model that could sustain itself? Does the math work, even at small scale?
Is AI live in production, making key decisions or doing core work? Could this product exist without AI, and if so, why does AI make it better?
Does the solution meaningfully move the needle in its chosen category? Is the problem real and the improvement noticeable?
Heralds collect evidence throughout the mission, not just at the end. The evidence pack is what the lab reviews at each gate and what determines whether the mission credential is earned.
Presented at day 30, day 60, and day 90 reviews.
Submitted as the final evidence pack on day 90.
The cadence keeps Heralds honest. Build during the week, review on Thursday, plan on Friday, rest on Sunday. The lab enforces this rhythm because sustained output beats heroic all-nighters.
Set the one goal that matters. Check Gemini pipeline health. Confirm the build target for the week.
Protected building time. Heralds work on the core product. No meetings unless something is on fire.
Talk to at least one real user. Test assumptions. Capture feedback. Update the evidence file.
Heralds present progress to the lab. Show working code, usage data, and blockers. Get feedback and direction.
Ship what was built this week to production. Write the weekly evidence note. Plan the next week.
Optional build time for Heralds who want it. Otherwise, step away. The mission is a marathon, not a sprint.
No build. No Slack. Read, reflect, or do nothing. Come back sharp on Monday.
Herald Labs Missions are open to Heralds in the current cohort. If you are not yet a Herald, the builder intake is the first door.
Pick your mission category, scope the hypothesis, and commit to the 90-day clock. The lab will assign a review partner and stack access.
Accept a missionThe builder intake is how you join the lab and become eligible for missions. Heralds are selected on evidence of building, not credentials.
Apply for builder intakeHerald Labs Missions exist to make AI business building less about pitch decks and more about working products with real users.