GeezerVison

The World Wide Web, restored to 800 pixels of glory.

GeezerVison is showing unify.ai — API · unify

API · by invitation

Keep your harness.Own the model that drives it.

Your agent improves itself from its own work, round after round: better, faster and cheaper the more it runs, in a model you own.

Request access

Better, faster and cheaper, the more it runs.

3×

the tasks solved, compared with its harness’s own memory

Its harness’s own memory24%

Trained by unify73%

Continual-ARC

1⁄2½

the steps taken per task

Its harness’s own memory12.9

Trained by unify6.8

Continual-ARC · steps of a budget of 16

1⁄3⅓

the GPU time to learn new facts without losing old ones

Standard LoRA22.6 h

Synaptic LoRA7.8 h

Synaptic LoRA · one H200 · research preview

Continual-ARC Why we built Continual-ARC

One base URL.

Point your harness’s model calls at unify and add one header. Your prompts, tools and provider key stay as they are, and every call becomes something your model learns from.

Python

from openai import OpenAIclient = OpenAI(    base_url="https://api.unify.ai/openai/v1",    default_headers={"X-Unify-Key": UNIFY_KEY},)

Recursive self-improvement

RSI. RSI. RSI.

01CaptureYour agent’s history first, then its live calls, through one base URL.
02ImproveYour users’ feedback, hindsight and frontier models judging every step show your model what to do better.
03ReleaseNew weights take your calls once they’re within the margin you set. Hold or roll back at any time.

The harness becomes its training pipeline.

Today an agent learns in its harness, as notes, skills and memory files that only grow. Here, what it learns moves into the weights, round after round, and the harness stays thin.

Your agentunchanged except for one URL: no reward to define, no labels to write, no environment to build

Your weightsserved for you, or yours to run anywhere, on-premises included

Your networkself-host the gateway and the training, and your calls stay inside it

The first round is on us.

We work with a few teams at a time. Tell us about your agent, and if it’s a good fit, we’ll train its first model on its own history and show you how much faster and cheaper it gets, before you pay anything.

Request access

You are visitor 01402503 — GeezerVison does not store the pages you look up.
Best viewed with Netscape Navigator 4.7 or Internet Explorer 6 at 800×600. No cookies required.