# Alot Labs

> Alot Labs is a founder-led applied-AI engineering lab. We design, build, and run
> AI systems ourselves — multi-agent orchestration, self-healing test automation,
> multimodal and knowledge retrieval, autonomous outreach pipelines, customer-facing
> agents, and safety-critical systems with guardrails enforced in code — then deploy
> them beside client engineering teams, in the client's cloud, with full handover of
> code and documentation. Seven systems built and run at the lab's own expense, each
> with a full case study.

Engagements start with a 45-minute technical architecture briefing (no pitch deck,
client engineers in the room). Contact: hello@alotlabs.com · pod applications: careers@alotlabs.com · security: security@alotlabs.com

## Pages

- [Deployments](https://alotlabs.com/deployments/): seven AI systems the lab built and ran end to end, each with a full case study
- [The lab](https://alotlabs.com/lab/): eleven projects on the bench — STG (the successor to AutoTest), nearby in phase zero, flywheel (personal wealth OS), holymolypoly (lag arbitrage, paper-only by rule), jaanal (a controlled multi-agent world that simulates sport play by play to predict games and matches — the MiroFish swarm-prediction thesis under lab conditions, aimed at a fine-tuned model of our own), LPP (a company as software, with a constitution enforced in code), academy (training and certification for agent engineers), RAW (a venture run as an AI-native skill stack), banks (a working bank for AI agents), compliance (camera-privacy layer, spec in review), and autoloop (the goal/verify loop kit every pod inherits)
- [Capabilities](https://alotlabs.com/capabilities/): engineering practices — retrieval/RAG, multi-agent systems with enterprise controls, security and safety invariants
- [Services](https://alotlabs.com/services/): engagement shapes, each with a defined output and handover
- [The method](https://alotlabs.com/method/): how an engagement runs, discovery through handover, with an exit at every stage — written to be printed, and the page /process/ now redirects here
- [Pricing](https://alotlabs.com/pricing/): how engagements are priced
- [About](https://alotlabs.com/about/): the lab and the people behind it
- [Careers](https://alotlabs.com/careers/): per-project build pods staffed with Claude-certified engineers — the Core, build pods, the Bench
- [Live](https://alotlabs.com/live/): live telemetry from four systems Alot Labs automates and runs at its own expense (audiolab, videolab, reactorlab, nearby) — every number exported by the system it describes, stamped with when it last ran
- [The standing commitment](https://alotlabs.com/commitment/): a written promise about what happens to a system after handover, and if Alot Labs stops operating — no service in the request path, a licence that survives, no kill switch, a published wind-down
- [Writing](https://alotlabs.com/writing/): two essays — what broke inside systems the lab ran at its own expense, and what a forward-deployed engineer actually does
- [Compare](https://alotlabs.com/compare/): what a big-four AI practice, two senior hires, and a contract build team each cost over the same weeks — a calculator with editable rates (published 2026 market ranges as defaults, sources listed on the page) plus, for each of the three, four cases where that option beats us
- [Contact](https://alotlabs.com/contact/): book a briefing or send a workflow description

## Lab entries

Each of the eleven lab projects has its own page, and each carries one of six fixed
states: Spec, Design, Building, Prototype, Running, Parked. A dated log of what
changed is at https://alotlabs.com/lab/#log — entries are copied out of each
project's own build log, not written for the site.

- [STG](https://alotlabs.com/lab/stg/) (Running) · [nearby](https://alotlabs.com/lab/nearby/) (Running) · [flywheel](https://alotlabs.com/lab/flywheel/) (Prototype) · [holymolypoly](https://alotlabs.com/lab/holymolypoly/) (Parked) · [jaanal](https://alotlabs.com/lab/jaanal/) (Prototype) · [LPP](https://alotlabs.com/lab/lpp/) (Building) · [academy](https://alotlabs.com/lab/academy/) (Design) · [RAW](https://alotlabs.com/lab/raw/) (Running) · [banks](https://alotlabs.com/lab/banks/) (Prototype) · [compliance](https://alotlabs.com/lab/compliance/) (Spec) · [autoloop](https://alotlabs.com/lab/autoloop/) (Running)

## Case studies

- [swarm](https://alotlabs.com/deployments/swarm/): multi-agent research with a live audit trail — human spend gate, 100% source re-verification, a writer with no web access
- [AutoTest](https://alotlabs.com/deployments/autotest/): self-healing browser QA from a one-line prompt — two-model assertion consensus, cached selectors, healing never touches assertions
- [sainma](https://alotlabs.com/deployments/sainma/): multimodal video search — three fused signals at scene granularity, five specialist agents, answers as watchable clips
- [nearby](https://alotlabs.com/deployments/nearby/): autonomous territory-to-meeting outreach — five LLM judgment calls, deterministic everything else, ~75% model-spend reduction
- [frontline](https://alotlabs.com/deployments/frontline/): 24/7 customer-facing agents — two personas, token-streaming chat, 100% of conversations auditable
- [corpus](https://alotlabs.com/deployments/corpus/): one workspace for team knowledge — five stores, one search, one generated schema
- [Meai](https://alotlabs.com/deployments/meai/): safety-critical AI — zero code paths can say "safe", deterministic dietary filters, confidence-gated vision