Hi! I’m Clay. I build useful software, explore hard problems, and write down what I learn.

Precise, cited math for drone mapping, surveying, aviation, and geodesy that runs on your device, even offline.

Thousands of clinical calculators, drug references, and billing codes behind one search box. Free, private, and fast enough to use at the bedside.

Practice the codes you cannot practice on patients. Full clinical simulations that run with no cost, no login, no install, and no network.

Every calculation a trade needs, cited and correct, in a browser that still works when the job site has no signal.

See what your paycheck, taxes, debt, and retirement actually look like without handing your finances to a single server.

Encrypt, sign, hash, split a secret, scrub metadata. A full crypto toolkit that runs inside your tab and uploads nothing, ever.

Read a contract the way a linter reads code. Finds the clauses that cost you later and cites the exact line behind every one.

Your AI assistant should answer to you, not to advertisers. Strips ads and affiliate tracking out of ChatGPT, Grok, and Perplexity, and catches hidden prompt injection.

A duel decided by reading your opponent, not by luck, grinding, or timers. Matches end in minutes.

Compute is the commodity and the market is the board. Corner the AI stack before your opponent does.

Your AI agents already hold your SaaS credentials. Proxilion makes sure they can only do what a human actually scoped, and proves it on every single call.

The last line between a model and a machine that can hurt someone. Fails closed, in hardware, before an unsafe command ever reaches the actuator.

Your coding agent stops guessing. One call tells it exactly what a task touches, one gate tells it what it must not break. No API key, same answer every time.

Your agent broke in production and the logs will not say why. Rewind the run, fork it at the exact step that went wrong, and prove the fix before you ship it.

The cloud posture platform that normally costs six figures, rebuilt in the open. 300+ policies across AWS, GCP, and Azure in YAML you can actually read and change.

Ask your logs a question in plain English and get a real answer. A serverless SIEM that runs for a fraction of what the incumbents charge.

Audit season without the scramble. Pulls evidence from 30+ platforms, shows exactly which controls are still short, and exports OSCAL that any GRC tool can ingest.

A leaked API key should not be game over. Qiuth turns bearer keys into proof-of-possession tokens that need more than the key itself to work.

A private AI that reads your documents on a laptop with the wifi switched off. Local models, encrypted at rest, free forever.

A signed safety boundary between an AI policy and anything that can move, dose, or open a valve. Refuses unsafe commands and leaves a verifiable audit trail even when the model itself is compromised.

Find out whether a robot dataset is safe to train on before it quietly wrecks your model. One command, any format, and a signed trust certificate you can hand to anyone.

Describe a part and get the bending, buckling, resonance, and bolted-joint checks back with the code clause behind every number. No cloud, no account, no silent pass.

About half of published biomedical models cannot be reproduced from their own papers. Reprolith rebuilds the model, re-runs it, and certifies result by result whether it holds.

Variant in, corrective edit out. CRISPR design across nuclease, base, and prime editors, with calibrated uncertainty and population-aware off-target nomination.

Most AI guardrails assert that they are complete. Verisim proves whether yours is, returning the exact commands it lets through, each one demonstrated against a real shell.

The tumor-growth and exposure-response models oncology drug development runs on, annotated by evidence strength and exported straight into NONMEM and SBML.

Anesthetic PK/PD parameters that carry their own applicability envelopes, so the model tells you when you have left the data it was built on.

Cardiac ion-channel block data that answers with a proarrhythmia risk distribution, and how often the safety call flips, instead of a confident verdict it cannot support.

Human gestational physiology, every parameter tier-annotated by evidence strength and exportable to SBML, CellML, and PhysioCell.