Skip to content

The research assistant
that can't fool itself.

An autonomous ML research harness for Claude Code. Turing runs the experiment loop (hypothesize, train, evaluate, decide, record) while you decide which experiments are worth running. Named after the person who first asked whether machines could think, then built the framework for answering the question.

"What comes across as luck is really the refinement of taste: years of feeling faint differences in the weight of the metal, the subtle offsets of a mis-mint." - Amy Tam

claude code
/turing:init # Scaffold project
/turing:train # Run experiment loop
/turing:brief # What happened?

Campaign: 8 experiments, 5 kept
Best: accuracy 0.87 (exp-005)
Exhausted: hyperparameter tuning
Recommendation: try feature engineering

/turing:try add polynomial interactions
/turing:train # It follows your lead

You bring the taste. Turing handles the discipline.

ML research is a search over a high-dimensional space of models, hyperparameters, features, and preprocessing strategies. The tragedy is not that this search is slow. It’s that when humans do it manually, the process is structurally unsound. Researchers modify training code and evaluation code in the same session. Results live in notebook cells, not structured logs. Promising directions get abandoned because someone forgot what they tried three hours ago.

Turing formalizes the experiment loop so that discipline is the default, not an act of willpower. The evaluation harness is immutable: the agent that trains models literally cannot see how they’re scored. Every experiment is logged. Every hypothesis is tracked. Convergence is detected automatically. The agent can run 700 experiments in two days with the kind of consistency no human can sustain.

Your job shifts from “person who types hyperparameters and reads loss curves” to “person who decides which hypotheses are worth testing.” From coin-flipper to coin-selector. That’s the division of labor that six decades of philosophy of science says is correct.

You Research taste Problem selection Judgment under ambiguity
/turing:try
Turing Hypothesize → Train → Evaluate Decide → Record → Converge? Immutable evaluation. No amnesia.
/turing:brief
Briefing Campaign summary • Best model What’s exhausted • What’s promising Recommendations • Next steps
↑ You again
🧪

Autonomous Loop

Hypothesize, train, evaluate, decide, record. The agent runs experiments while you sleep. /loop 5m /turing:train for fully hands-off operation.

🔒

Immutable Evaluation

The agent can’t see evaluate.py. Six defense layers prevent metric gaming. Every prompt-based rule got worked around; every code-based rule held.

🧠

Structured Memory

Hypothesis database with priority queue and detail files. Novelty guard blocks duplicates. Experiment families with exhaustion detection. Nothing forgotten across sessions.

🎯

Taste-Leverage Interface

/turing:try injects your judgment. /turing:brief reports intelligence. The loop is bidirectional by design: human taste steers, agent discipline sustains.

📊

Statistical Rigor

Multi-seed studies with confidence intervals. Reproducibility verification. Calibration checks. Regression gates. Stop publishing lucky seeds. Publish distributions.

⚙️

Two Specialized Agents

@ml-researcher modifies code and runs experiments. @ml-evaluator analyzes results read-only. An analyst who can’t act on observations makes more trustworthy observations.

Inspired by Karpathy’s autoresearch, AutoCrucible, TreeQuest, Popper’s falsificationism, Goodhart’s Law, and the principle of least privilege.