cd ../
cat manifesto.txt | less
____  ____   ___   ___  _____   _____ ___  _   _ _   _ ____  ______   __
|  _ \|  _ \ / _ \ / _ \|  ___| |  ___/ _ \| | | | \ | |  _ \|  _ \ \ / /
| |_) | |_) | | | | | | | |_    | |_ | | | | | | |  \| | | | | |_) \ V /
|  __/|  _ <| |_| | |_| |  _|   |  _|| |_| | |_| | |\  | |_| |  _ < | |
|_|   |_| \_\\___/ \___/|_|     |_|   \___/ \___/|_| \_|____/|_| \_\|_|
════════════════════════════════════════════════════════════════════════
$cat 00_the_problem.txt

> ML research is broken.

Notebooks rot in forgotten directories. Experiments vanish into unnamed folders. Git commits say "asdf" and "final_final_v3". Results that took weeks to produce become impossible to reproduce.

The people with the best ideas lack the infrastructure to execute. The people with the infrastructure lack time for real questions.

$cat 01_thesis.txt

> You do not need 100 researchers.

You need judgment. Ambition. A system that multiplies effort instead of drowning it in operational overhead.

A few engineers with the right tools can operate like a frontier lab.

$cat 02_how_it_works.txt

> Autonomous agents handle lab work.

┌─────────────────────────────────────────────────────────────────┐
│  AGENT LOOP                                                     │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  [1] Plan experiment based on research memory                   │
│       │                                                         │
│  [2] Generate code, track all artifacts                         │
│       │                                                         │
│  [3] Hit approval gate for expensive compute    ◀── YOU DECIDE  │
│       │                                                         │
│  [4] Execute with full provenance                               │
│       │                                                         │
│  [5] Record evidence chain: code › data › metrics › claim       │
│       │                                                         │
│  [6] Update research memory for future runs                     │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

Every claim links to code, data, config, and logs. Nothing lost. Nothing hand-waved.

$cat 03_human_control.txt

> Humans govern every decision that matters.

Agent does not spend $10k on GPU without asking. Does not publish without review. Does not modify production data without approval.

┌───────────────────────────────────────────────────────────────┐
│  APPROVAL GATE                                                │
├───────────────────────────────────────────────────────────────┤
│                                                               │
│  Agent requests: gpu_execute                                  │
│  Estimated cost: $2,847.00                                    │
│  Duration: ~4 hours                                           │
│                                                               │
│  Rationale: Fine-tune Qwen-2.5-72B on curated dataset         │
│  to beat GPT-5 on HumanEval+ benchmark                        │
│                                                               │
│  [y] Approve    [n] Reject    [?] Details                     │
│                                                               │
└───────────────────────────────────────────────────────────────┘

You stay in the loop. Agent stays productive.

$cat 04_for_whom.txt

> Built for cracked engineers who want to do real research.

-- PhD students tired of losing track of experiments

-- Indie researchers who want lab-grade infrastructure

-- Small teams competing with orgs 100x their size

-- Anyone who believes great work needs great tools

$cat 05_the_future.txt

> Research should compound, not decay.

Every experiment teaches the system. Every failure narrows the search. Every success becomes foundation for the next question.

The lab remembers everything. So you can focus on the science.

════════════════════════════════════════════════════════════════════════
┌─────────────────────────────────────────────────┐
│                                                 │
│  Proof Foundry alpha                            │
│                                                 │
│  "A few engineers can now operate               │
│   a serious AI lab."                            │
│                                                 │
│  EOF                                            │
│                                                 │
└─────────────────────────────────────────────────┘