dojo7

If you check out the post frequency on the website, you’ll notice around 18 months back, I started clearing a decades-long backlog and posting a lot.

This post is the third in a multi-part series called What We Owe Machines.

This post is the second in a multi-part series called What We Owe Machines.

This post is the first in a multi-part series called What We Owe Machines.

This post is the companion to a preprint, Manufactured Rank Is Real but Unrealized: An Optimization Barrier in Nonlinear Low-Rank Adapters for Task Adaptation.

Charles nailed it The shopping center at Noosa civic has a small table that sits just outside the supermarket.

Every wallet in this post was drained years ago.

This experiment is Part III of a three-paper series: Part I — The Amortization Envelope: Where a SHA-256d Mining Advantage Can and Cannot Live: mining cannot beat brute force unless SHA-256 is broken.

This post walks two parts of a three-paper series: Part I — The Amortization Envelope: Where a SHA-256d Mining Advantage Can and Cannot Live: the reduction and the bounded sharing envelope.

Part III leaves the series in an unfamiliar position: holding a real result.

This is the third post in Tilting at Windmills, a series where we take machine learning approaches, point them at the Bitcoin mining problem, and examine how they fail.1 Part I shows that the expected gradient through double-SHA256 is exactly zero, so not…

In 2025, almost every AI automation framework was built on the same model: autonomous agents, sending messages to each other.

In October 2025 Anthropic put out a paper on whether language models can introspect.1 It’s careful work with a careful set of caveats, and within about a day of it landing I started seeing the caveats thrown overboard.

No indie website is complete without bad poetry.

Part I of this series applies REINFORCE to nonce prediction and finds that the expected gradient is exactly zero: SHA-256’s avalanche effect makes the score of a sampled nonce independent of each of its bits, so every parameter update is averaged noise.

Code on GitHub at captainpete/keycarver.

Code on GitHub at captainpete/earnest.

The full paper, with methodology and results, is available here.

This is the first post in a series called Tilting at Windmills, where we take machine learning approaches, point them at the Bitcoin mining problem, and examine how they fail.1 Bitcoin mining is useful for this because it is genuinely unsolvable by any me…

The full paper, with methodology and complete results, is available here.