Tempered Labs — I build AI software for other companies, and use that revenue to build products I co-own.
The studio of Tyree Pearson, founder-engineer, New York
Hard problems, carried to production.
I build for clients, and I build my own.
The consulting funds the ventures. The ventures prove the consulting. Everything I sell, I have built for myself first.
Advisory for businesses under stress.
Operating businesses hire me to unstick a stalled roadmap, rebuild the architecture, or put AI into daily operations.
The roadmap has no owner.
I take the CTO seat and set the technical strategy: the product roadmap, the architecture, the stack, and where AI fits. Each recommendation is a written decision with its reasoning, so your team can act on it without me.
The product is built, and will not finish.
The last 20 percent is rarely features. I design it, build it, ship it, and hand it over running, with acceptance criteria on every milestone.
There is no engineering team, or the wrong one.
I stand up the crew from a contracted vendor, set the cadence and the standards engineers stay for, then hand the team to you and step back.
The AI pilot never made it into operation.
I put agents into the stack with named goals and a named owner. If nobody owns it, it is not in production.
What the company knows leaves when people do.
I build the company knowledge base: one place every agent feeds and every decision draws from, with a documented schema and a machine-readable change log.
Carry it back.
Hard problems, carried to production.
I pressure-test the idea before committing hours, then build it to ship.
I build, operate, and co-own software ventures.
Each venture runs its own AI agents; all of them write to one shared knowledge base. Each venture gets its bar set in writing before I commit hours.
Four ventures, one forge.
Guessers, a private prediction game for friend groups. Puggy, an AI notes companion. Ruleshift, a compliance workspace that shows its work. PatientLXA, a provider-facing bone-health brief.
Different stages, one standard.
One in production, one in pilot, one in closed beta, one on the waitlist. Each bet is judged on its own numbers, against a bar set in writing.
A private prediction game for friend groups.
Built and run by the studio. Closed beta; waitlist open at guessers.app.
An AI notes companion.
Puggy is an AI notes companion, co-built with the engineer who wrote it. The waitlist is open.
A compliance workspace that shows its work.
Co-founded by Tyree Pearson with Alec Brady and Austin Flowers. Your team uploads evidence; named reviewers publish requirement-by-requirement assessments. In production.
A provider-facing bone-health brief.
Operated by Patient LXA LLC. It assembles a patient's bone-health record before the visit, for the provider. In pilot.
Carry it back.
Bring the problem your team is stuck on.
Fixed fee, not hourly. Written acceptance criteria on every milestone. You own the IP from day one.
Tempered Labs · Advisory
The work, by the problem it solves.
Insights engagements are delivered by Tyree Pearson under Tempered Labs. I take few clients, with deep involvement, and I do not sell a capability I have not carried myself.
Fractional CTO
I take the CTO seat and set the technical strategy: the product roadmap, the architecture, the stack, and where AI fits. Each recommendation is a written decision with its reasoning, so your team can act on it without me.
At a US membership community the technical lead took leave, and the roadmap went with him. I took the CTO seat. The entry was an audit: 3 production systems, a risk register with a pass/fail bar per system, and a definition of live for each, written before any fee. All 3 shipped signed off. The seat is now a monthly advisory retainer: the product direction and the stack decisions come through me, each written with its reasoning so the CEO can act on them without me. Ongoing. As of September 2026.
Platform building
The last 20 percent is rarely features. I design it, build it, ship it, and hand it over running, with acceptance criteria on every milestone.
I have taken a consumer product from build to production, shipped a mobile app through TestFlight, and moved an operations platform from V1 to V2 on Postgres, migration included. At the same community, I rebuilt an operations-to-CRM pipeline whose scheduler ran 35 versions behind, and held release until 7 consecutive syncs ran clean, so the revenue team works from real data. Every live change stays staging-first, dry-run, and reversible. Ongoing. As of September 2026.
Engineering leadership
I stand up the crew from a contracted vendor, set the cadence and the standards engineers stay for, then hand the team to you and step back.
I run a contracted engineering team across two countries, billed through one vendor. Security work is split between a builder and a certifier, so no engineer signs off on their own code. At the membership community, I staffed 3 engineers through the vendor, vaulted the mobile signing keys, and documented the build, so releases no longer rest on one person's access. I write the audits so the next owner can run them without me. Ongoing. As of September 2026.
Operational AI
I put agents into the stack with named goals and a named owner. If nobody owns it, it is not in production.
For a wound-care and regenerative-therapy distributor, I built two assistants on one engine behind two policies: one internal, one for partners in shared channels. The external assistant answers only from an approved source and holds no write tools; its channel allowlist fails closed, and every exchange appends to an audit log. My own stack runs the same discipline: a library of agents enforces engineering standards instead of describing them, and a reconciler closes an issue only when a merged pull request proves the work shipped. Ongoing. As of September 2026.
One shared brain
I build the company knowledge base: one place every agent feeds and every decision draws from, with a documented schema and a machine-readable change log.
That engagement began as discovery: what the company knows, where it is kept, and how it is used to make decisions. It became one canonical workspace that syncs out to the documents partners read and pulls their comments back, with a scheduled auditor that reports drift without ever editing anything. The studio runs on the system I sell: every asset carries its own metadata, so the library stays searchable as it grows. Ongoing. As of September 2026.
How an engagement starts
- A first call: what broke, what it is costing, what you have already tried.
- An audit of what already runs, with a pass/fail bar per system.
- A definition of live for each system, written before any fee is agreed.
- One fixed fee against written milestones.
- Handover: documentation, keys, and the record of every call. Then I step back.
How I work
- Fixed fee, not hourly. You know the price on day one and I carry the overrun.
- Written acceptance criteria on every milestone.
- You own the IP from day one, in writing.
- Scope fenced in writing. What is excluded is named, so later requests get priced instead of absorbed.
- Few clients, deep involvement. No enterprise procurement, and I do not staff bodies.
- After an engagement, an advisory seat is optional, at a flat monthly fee.