01Closed beta
Leverage
Less trading bot than proving ground: autonomous systems in a hard domain.
Leverage
Structured read
Session context
Shadow learning
Live chart
Market map
Pulse scan
Floor
Co-Pilot
Playbook
Inputs
Desks
Blueprint Labs
Research that
ships.
Systems for hard domains. Evidence before theater.
Building
Things we're building
Real systems in motion. Each one a slice of the lab made tangible.
02Research
ACRA
Auditable clinical reasoning. Evidence in, grounded steps out, with a paper trail you can inspect.
ACRA
Ground claims
Bind sources
Trace steps
Evidence
Retrieve
Reason
Audit log
Corpus
Trace
03Open Beta
Apex
Solana trading with AI in the loop. Surfaces calls and can execute when the setup clears internal gates.
Apex core
Detect
Evaluate
Decide
Chain event
Signal read
Verify
Risk gates
Execute
Ingest
Act
04Early Access
/Slash
A repo-aware engineer for GitHub and VPS: chat, edits, reviews, and live monitoring from one mobile-first workspace.
/Slash agent
Repo context
Prompt to edit
Live ops
GitHub
VPS / SSH
Mobile
Edit / PR
Review
Monitor
Sources
Actions
05Open Beta
Pulse
GitHub noise, compressed. Short digests of what actually changed for teams that live async.
Pulse
Watch repos
Compress noise
Ship digest
Commits
Pull requests
Digest
Async read
Activity
Signal
07Experiment
Lexa
Legal-domain research on case law and long threads. Answers that stay tied to the corpus.
In experiment
Lexa
Retrieve
Weigh
Stay faithful
Case law
Firm notes
Ground
Draft
Corpus
Output
Stack
Research in.
Products out.
Agents, models, and tools share one core. That is how a small lab ships more than demos.
- 1Study a real problem until it cracks
- 2Reuse the same agents, models, and infra
- 3Ship a product people can actually use
Blueprint core
Shared infra
Evals
Playbooks
Research
Agents
Models
Data
Leverage
/Slash
Pulse
Apex
The lab
Shipped
Focus
Where we spend time
We bet on a few directions and build real things in them. Not papers, not decks.
01
Agents
Production agents with traceable reasoning. Built to do real work in real systems. Not to impress a demo.
02
Fine-tuned models
Domain-specialized models trained on narrow slices of the world: law, medicine, code. We care about what models actually know.
03
Workflow systems
Automation with humans in the loop. We study the real work before we replace any of it.
04
Developer tools
AI tools for engineers. Most of them end up being things we wanted to exist for ourselves.
The lab
Hard domains deserve systems that ship.
We study real problems, build proving grounds, and open products when the evidence is there.