ask · 104 questions · cited answers

The questions your system
can't answer.

Two hundred of them, grouped by the work you're trying to do. Every answer cites its source and refuses to guess when the evidence is not there. This is what the model is for.

six buckets · 6 visible·read by humans, agents, scripts
§01 · what just broke

What just broke. And what caused it.

Health, regression, drift. The on-call brain that already had context before the page fired.

What is the health status of the fleet?

What just went degraded?

What caused billing's p99 to spike at 14:02?

What caused this memory leak?

Which commit broke the auth tests?

Show me every change between the last green run and the first red.

Was anything else degraded at the same time?

What is the smallest revert that fixes this?

if this is the kind of morning you keep having

Five agents already run on the model.

Nightly triage. Blast review. Incident first responder. Cost cartographer. Drift watcher. Each one cites its evidence and refuses to guess.

See the five agents in action →
§02 · what breaks if I touch this

What breaks if I touch this.

Every change has a blast radius. The model walks it before the PR ships, so the conversation about risk happens in the diff, not in the postmortem.

What would break if I change this service?

If I land this PR, who needs to be ready for it?

What is the blast radius of bumping this library?

Can I delete this method? Who is still calling it?

Can I rename this endpoint? Are there observed consumers on the old shape?

What is the safe ship order for these five services?

Can I roll this back without breaking a consumer?

§03 · where the money and the rot go

Where the money goes. Where the rot goes.

Cost, zombies, capacity. Spend mapped over the graph instead of the billing CSV. Dead code surfaced before it rots into someone's morning.

How much does this service cost per month?

What is our top cost driver this week?

Which services have no consumers?

Which endpoints have not been called in 30 days?

Which database tables haven't been read in 90 days?

Which jobs cost more than they save?

Why did our bill jump 18% week over week?

Which feature flags are 100% on or 100% off?

§04 · who owns this and who knows

Who owns this. Who else knows.

Ownership without the spreadsheet. The team, the on-call, the last engineer to touch it, the de facto expert nobody officially named.

Who owns this service?

Who is on call for billing this week?

Who was the last engineer to touch this file?

Where is the repo for this service?

Who has touched this code in the last six months?

What does this team own?

Who do I escalate to if I cannot reach the primary on-call?

What systems should I look at first to understand checkout?

if you're the architect everyone pings

Stop being the only person who can answer.

Cohort zero has five seats remaining. Founding-customer pricing forever, weekly contact with the people writing the code.

See the design partner program →
§05 · is this safe to ship

Is this safe to ship.

Security surface, compliance posture, provenance per edge. The receipts that turn a confident answer into one a human can actually defer to.

What is the attack surface of this service?

Which endpoints are exposed to the public internet?

Which services hold PII?

When was the last security scan? What did it flag?

Are there observed consumers on the old shape?

How sure are we about this edge?

Where does this fact come from?

When was this last verified?

§06 · what agents do with it

What agents do with it.

The model is exposed to Claude, Cursor, Cline, and any in-house agent over MCP. Agents inherit the same discipline humans get: cite or refuse.

What is the safest agent action I can take here?

What does the agent know about my system before answering?

Show me the receipt for this verdict.

What is the confidence of this answer?

Why does the agent refuse to guess when it does?

How can my agent walk the blast radius of a change?

show all 59 more questionsexpand ↓
topology
  • Who calls this endpoint?

  • What does this service depend on?

  • Which services consume this event or topic?

  • What database does this service write to? Read from?

  • Which services share a database, table, or row?

  • Who reads this column?

  • Show me every consumer of this internal API.

endpoints / contracts
  • What endpoints does this service declare?

  • What is the schema of this endpoint?

  • Has this endpoint's contract changed?

  • What is breaking about this contract bump?

  • Is this endpoint behind the WAF?

  • What is the rate limit on this endpoint?

events / queues
  • Who publishes to this topic?

  • Who subscribes to this topic?

  • What is the consumer lag on this queue?

  • Is anything dead-lettered?

  • What event triggered this side-effect?

build / ci / images
  • What image is running in prod for this service?

  • What was the last build status?

  • What is the test coverage on this repo?

  • Are there builds older than 30 days still running anywhere?

  • Which images have known CVEs?

multi-environment
  • What is running in prod-eu vs prod-us?

  • Is staging drifted from prod?

  • Why is fraud-ml on v412 in prod and v410 in staging?

  • Which services are out of sync between environments?

  • Where is this incident reproducible?

drift
  • What edges does the schema say exist but runtime has never seen?

  • What edges does runtime see that no schema declares?

  • Where does the manifest say one replica and Kubernetes is running five?

  • Has this service drifted off its declared CPU or memory request?

  • What changed silently outside of CI?

releases / planning
  • What is the safe ship order for these five services?

  • Can I roll this back without breaking a consumer?

  • What is the smallest deploy that finishes this migration?

  • Are any services blocked by an upstream that has not shipped yet?

onboarding
  • What does this team own?

  • What does this service do, in one paragraph?

  • What systems should I look at first to understand checkout?

  • Where is the runbook for our on-call rotation?

  • Where is the dashboard I should keep open?

integrations
  • Which Stripe webhooks do we listen to? Which do we send?

  • Which services hit Snowflake? With what query shape?

  • Which services use Auth0?

  • How much do we send to PagerDuty per week?

  • Is any integration drifting from its SLA?

compliance
  • Which services touch PII?

  • Which services are inside the PCI scope?

  • Who has accessed this service's secrets in the last 30 days?

  • Where is the data residency boundary for this customer?

performance
  • What is the p99 latency of this endpoint?

  • Where does the latency budget go on this request path?

  • Which downstream is the slowest in this trace?

  • Has this endpoint's latency profile changed in the last week?

capacity
  • What is the headroom on this service right now?

  • Which services need to scale before our next traffic spike?

  • Which services are over-provisioned and could be shrunk?

  • Where will we hit a DB connection cap first?

how we ask

Three surfaces, one source of truth. Assistant in-app for humans, MCP for agents, API + SDK for code. Same nodes, same edges, same answer.

The agent API →
why we built this

The map versus the territory. Observability is the territory at max resolution. blastmap is the deliberate abstraction with stable coordinates and confidence on every edge.

The manifesto →
why we are raising

Build the substrate first. Agents fall out of it cheaply. The data network effect compounds across customers. Seed round open.

The seed round →
pre-launch · v0.1 incoming

Ship without fear
of the unknown.

Get the v0.1 launch notification and a 30-day head start before public release.

we'll only email you twice. no marketing.