Grade Engine — sample mode · live scan rolling out Evaluation-led · Web · App · Web3
Houtos Labs
Start a conversation

The work — labeled exactly as it is.

No invented clients. No inflated numbers. A live product you can use right now, a hand-built interactive demo, an engine mid-build, and lab concepts marked as concepts — because a studio that grades other people's honesty has to lead with its own.

2 live · 1 in build · 2 concepts Status verified 2026-07-02

W-01 · Evaluation platform · Live

The Grader

A live audit instrument: six lenses, one honest grade, and the fix list that moves it.

Open the project file
Challenge
Development firms guess at quality, and buyers have no instrument to verify claims — so trust defaults to whoever pitches best.
Approach
Six-lens grading over 26 weighted metrics with confidence awareness. Every number is labeled Measured or Estimated — an estimate is never dressed as a verified result.
Outcome
Live on houtos.io. Grading real sites in sample mode today; the live-scan engine (Lighthouse + axe-core) is rolling out.
System type
Evaluation platform · scoring engine · report generator · plain-language synthesis
Use the live tool Sample report

W-02 · Analysis engine · In build

Peregrine Precision Evaluator

The measurement engine beneath The Grader — instrumentation replacing estimation, stage by stage.

Open the project file
Challenge
Preview-mode grading has an honest ceiling: without live scans, some readings must remain Estimated. The only structural fix is a real measurement backend.
Approach
A universal module contract feeding five analysis engines — accessibility (WCAG), privacy, EU AI Act, security, and repository quality — with Lighthouse + axe-core doing the measuring.
Potential
Flips Estimated readings to Measured across the board — same rubric, higher confidence, narrower fix-lift ranges.
System type
Analysis engine · crawler · module pipeline · AI synthesis
How the phases fit together

W-03 · Agent architecture · Concept — internal R&D

Autonomous Intelligence Desk

How far can an autonomous system be trusted — and what governance makes that trust rational?

Open the project file
Question
What does it take for an autonomous system to earn each level of authority, instead of being handed it?
Approach
A promotion ladder — research → simulated operation → reconciliation → limited live authority — where advancing requires measured performance, never enthusiasm.
Status
Internal R&D. Findings inform our client work on agent governance and evaluation harnesses.
System type
Agent architecture · risk governance · telemetry & reconciliation

W-04 · Memory architecture · Concept — internal R&D

Companion Systems

Engaging AI companions whose memory stays fully legible to the person they serve.

Open the project file
Question
Can a companion be genuinely engaging while its memory and personalization stay fully legible to the user?
Approach
Hybrid memory architecture, consent-first personalization, and safety layers designed before engagement mechanics — not after.
Status
Internal R&D. Findings inform our client work on AI product design and responsible personalization.
System type
Memory architecture · personalization · safety design

W-05 · Interactive demo · Live

The Hidden Lab

One hand-written file: a CSS-3D phone, canvas dashboards, and honesty labels on every simulated number.

Open the project file
Purpose
Show, don't tell — the front-end and systems range clients are buying, in a form you can poke at.
Approach
Everything drawn in code: Canvas 2D charts, a CSS-3D device, streaming chat mechanics. Honesty travels with it — Simulated and Paper-mode labels throughout.
Outcome
Live on this site. One file, 60fps target, reduced-motion and no-JS safe.
System type
Interactive demo · canvas engineering · motion system
Enter the hidden lab Tip: there's a secret inside — ↑↑↓↓←→←→BA

Your system could be next on this page.

Bring us the ambitious thing — the AI product, the protocol, the platform. We'll tell you what it takes to make it hold up under measurement.

Start a conversation Explore capabilities