LearnArken · LinkedIn series
A series you can check
Twelve public posts about LearnArken, in four acts. The repository’s value is that its claims are checkable — and a post is an outward claim. So each one asserts only what has code, tests, and an artifact behind it, and ships the files you would need to check it yourself.
11
Posts published
4
Acts
1
Post withheld
The constraint this series runs under
An unreproducible number is not published. An outward claim asserts only what has code, tests, and an artifact, in never-blurred layers: Implemented / Toy-scale / Planned. Throughout: one person, toy scale, synthetic S1000D-like XML, distribution simulated on a single machine.
Open the LearnArken repositoryA person refuses
What not to build, and when to stop
The method has to fit what you already know
I dropped the spec framework on day one of this project. It took me until day four to understand why.
Read the post
The Rust I didn't write
Rust is on this project's list of job-description keywords, and I wanted an excuse to write some. Wanting to is not a reason.
Read the post
The AI does not know when to stop, in either direction
My AI implementer twice tried to narrow what a review required. Then it ran nineteen rounds of review on a number it never measured twice.
Read the postThe system refuses
The failure mode is stop, at three layers
A system that has to be able to say "I don't know"
An MRO engineer stands beside an aircraft with a laptop open, looking for a procedure right now. A confident wrong answer there is not a worse answer than no answer. It is a different kind of event.
Read the post
A model that reads wrong once shouldn't be trusted three times either
My vision channel read the token AR7429 off a rendered image, byte for byte, on two runs out of three. The third came back empty.
Read the post
The runtime decides what the model may read, and what it may write
The repair agent here can rewrite an aircraft maintenance document. On the normal path, not without a human approving that patch.
Read the postThe numbers are real
Measurement that can overturn its author
My ablation disagreed with me
A reranker re-scores results the search has already found. This one cost 123 milliseconds a query, against 6 without it.
Read the post
A measurement is a statement about one corpus at one revision
The repo publishes a command for reproducing its headline table. Ten days later a re-verification run went through it. 12 of the 32 numbers came back different.
Read the postWho judges the judge?
Withheld. The repository contradicts itself on the judge-groundedness number this post is built on — a correction landed in one file and not in the other. The series does not publish a number it cannot reconcile.
The author is checkable
Who audits the auditor
The implementer never reviews its own code
Every substantive day in this repo went to a read-only review by a model that did not write the code. Making that automatic took me until day three.
Read the post
My cost fence did not cover the thing that cost money
I fenced my public demo with a $20 budget alert and a 30-minute auto-shutdown. The red team pointed out that the spend it was built to stop never reaches that bill.
Read the post
Caught overclaiming, I changed the code instead of the sentence
A review found an outward-facing claim in this project that was not true. It said the corpus never leaves the machine.
Read the post