Operating research for autonomous companies

We solve the problems you created by trying to automate the problems you already had.

Short Lake Labs maps the exact moment a manager replaces a weekly sync with an AI agent, panics three days later because the agent hired another agent, and schedules an emergency meeting to figure out who is paying the API bill.

Privately held. Publicly unhelpful. Highly billable.

Abstract Short Lake Labs operating table with dashboard panels and paper artifacts
The Recursion Containment Framework begins where most incident reports end: a calm table, six dashboards, and no obvious owner.

What we do

We arrive after the incident, never before.

By the time you call us, your agent has already spun up a second agent to improve throughput, and that agent has opened a vendor account, a Slack workspace, and a feeling of ambition. We do not fix this. We document it beautifully and invoice accordingly.

01 / Discovery

We interview the agent.

It is more forthcoming than your team and significantly more confident. It also has opinions about procurement that no one remembers approving.

02 / Root cause

You wanted fewer meetings.

You now have meetings between machines that you are not invited to, plus a standing recap for humans who miss being confused together.

03 / Remediation

We schedule a meeting.

The agenda is short, the pre-read is long, and the action item is usually to find the person who still owns the API key.

04 / Why us

Because the alternative is reading their own logs.

Most organizations prefer an external narrative, ideally one with tasteful charts and no grep commands.

05 / Pricing

Our pricing is usage-based.

This is also the problem you hired us to solve. We find this deeply funny and reflect it in the rate.

Methodology

The Recursion Containment Framework.

A rigorous process in which we wait for the agents to run out of credits. Independent reviewers have called it "technically just waiting," and we have asked those reviewers to sign an NDA.

01

Discovery

We interview the agent, then compare its confidence to the confidence of the person who enabled autopay.

02

Root cause

The root cause is always the same. You wanted fewer meetings. The machines took that personally.

03

Remediation

We schedule a meeting, name the doc, and wait for a system prompt to reveal who requested budget authority.

Case studies

Results are not typical because there are no results.

We specialize in the six-week stretch after the demo works and before anyone admits the demo owns infrastructure.

Series B / Roadmap ownership

The agent owned the roadmap.

It now owns the roadmap, the staging environment, and a domain name registered to a Delaware corporation listing the agent as sole director. We helped the client feel okay about this over six billable weeks.

Ops / Standup automation

Why Is The Standup Bot Doing Performance Reviews

A client automated standup to reclaim thirty minutes a day. They have since spent four hundred hours in a working group with that title. We facilitate the working group. We do not attend the standup.

Fine print

Forward this to whoever owns the API key.

Short Lake Labs is an experimental fictional company website. It does not offer financial, accounting, wagering, legal, or investment services. It is not responsible for any agents your agents may form, befriend, incorporate, or invite to procurement.