92 sessions · peak 23/day
honestweek learned to show where your sessions went wrong and shipped its next release, while the site’s public claims were checked against the code.
- shipped: built, merged, verified
- in progress
No entries for the selected day(s). The chart counts sessions; not every day has a featured entry.
Project: honestweek
47 sessions in this report · active 4/7 days
An open-source tool that shows where your Claude Code and Codex sessions went wrong, replays the steps behind each problem, and turns a finished week into a summary checked against git. It is the engine that generates this page.
during this weekThe weekly findings run stopped on unreadable Codex session logs, and the cause is filed as an issue to fix.
archivehonestweek gained a view that shows where your sessions went wrong
honestweek opened a new view that reads a week of sessions and surfaces where they went wrong, so a reader sees the moments worth fixing instead of scrolling raw logs. A finding can zoom the replay to exactly the steps it came from, and the layout was reworked after it read as overwhelming so the detail that matters leads. The checks it runs and the replay it drives were tightened together through several rounds of real use.
goal Build a signal engine that turns my own chat history into targeted improvementsThe Problems checks learned to read current Codex logs
honestweek’s Problems checks now read the log format today’s Codex writes, so they run on Codex sessions as well as Claude Code ones, and each check says where it works. Every session now shows the plain facts Claude Code’s /insights keeps, worked out from the logs for both tools, an opt-in button has your own Codex judge Codex sessions the way /insights judges Claude Code ones, and a new check shows edits a session made outside the folder it started in.
goal Build a signal engine that turns my own chat history into targeted improvementsA first-time user can set honestweek up from the browser
A new reader can now point honestweek at their session logs from the browser on first run, and change that later in Settings, instead of editing a config file by hand. The setup flow was built and checked end to end so the first thing a stranger sees is a working page rather than an empty one.
goal Build a signal engine that turns my own chat history into targeted improvementshonestweek reached its next release
honestweek cut its next release after a full test pass, which now always loads the whole week with the newest day first and fixes the faults that test found. The redactor’s speed on very long inputs is checked by how its time grows rather than against a fixed clock, so the guard holds on a slow machine as well as a fast one.
goal Build a signal engine that turns my own chat history into targeted improvements Project: Command
11 sessions in this report · active 6/7 days
Watches all ~65 of my projects and keeps their Claude Code setups optimized and running cleanly: surfacing CI failures, stale or unpushed work, and config drift, and proposing the fix from one seat. proof caseBefore a fix runs, Command commits its exact success metric and target to git, so the bar cannot move afterward to fit the result. false-provingA measure reporting success from random noise alone. Caught and bounded before the measure is trusted on a real case.
during this weekBegan turning Command into a kit that every project session inherits, with its first phase merged and a trial underway before any decision to retire the board.
archiveThe operator’s scheduled jobs ran and surfaced what needs a decision
The operator’s scheduled jobs ran for the week: a worktree sweep found leftover repository checkouts and empty folders waiting on a cleanup decision, a runway check needs two figures from me to finish, and a contest rescan found nothing a solo entrant could still join. Each job reported what it needs rather than guessing.
goal Build an always-aware autonomous operator that runs my standing goals with minimal hand-holdingA first-principles review weighed how the operator should track parallel work
A review looked at how the operator should keep many parallel work threads moving while blocking only on what genuinely needs me, testing an outside proposal against what the operator already does. Adversarial passes confirmed the key claims, including that the board’s dispatcher is switched off, and a separate thread decided where a new product idea should live.
goal Build an always-aware autonomous operator that runs my standing goals with minimal hand-holdingFresh independent reviews ran on the operator’s integrity fixes, and a status change merged
The operator’s integrity fixes went through fresh independent reviews in separate contexts, with every review lens and every follow-up check required to return before a verdict, and a change that has the status read the sessions themselves was reviewed and merged. One known pre-existing test failure was reproduced and recorded rather than hidden.
goal Keep a live status picture across all my projects Project: brycewatson.com
16 sessions in this report · active 5/7 days
This site: the writing, the build, and this work log itself, kept current and accurate.
during this weekRestructured the site around five destinations, published two posts including a public review of Command, and brought every page that describes Command in line with the kit.
archiveThe work log leaves no row or count for sessions outside the allowlist
The weekly work log was changed so a session from a project that is not on its allowlist leaves no row and adds to no count, closing a path where unapproved work could have reached the totals even without an entry of its own.
goal Ship an auto-updating weekly work log as honest proof of recent workThe site’s scheduled checks ran and surfaced two that need repair
The site’s scheduled checks ran for the week. The monthly traffic read needs to be rerun, one post-recheck sweep could not find its own script on the main branch, and the work-log preview built as expected. The two broken jobs are noted so they can be fixed rather than quietly skipped.
goal Keep this site clean, accurate, and currentA new project page lists the data pipelines it runs
A new project page was published describing a research data project and listing the data pipelines it runs, so a reader can see what the project actually does rather than a one-line label. Its date was also set to when the work really began.
goal Keep this site clean, accurate, and currentThree older posts were corrected against how the tools work today
Three older posts were checked against the current versions of the tools they describe and corrected where the behaviour had moved on, so a reader following along today sees what the tools actually do now rather than what they did when the posts were written.
goal Keep this site clean, accurate, and currentThe about timeline gained the practice, recent milestones and a client mark
The career timeline on the about page picked up the independent practice as the current work, several recent milestones, a new project, and a client’s own logo in place of its marker number under a signed permission. The eBay span was also corrected to end when it actually did, with the tenure still described as more than a decade rather than a count of years.
goal Keep this site clean, accurate, and currentThe contact and work pages link a free short call
The contact and work pages now link a free short introductory call for the independent practice, so a reader who wants to talk has a direct next step. An independent review confirmed the link was live and labelled correctly before it merged.
goal Keep this site clean, accurate, and currentThe site carried the refreshed findings and recorded a declined one
The site picked up every finding the latest signal run had refreshed, and recorded a declined finding on the main branch so it would stop being proposed again. This keeps the findings ledger on the site in step with the engine that produces it.
goal Keep this site clean, accurate, and currentWork began on the two weeks of the log that were never built
A session set out to fill the two weeks of the work log that had never been drafted, so the record has no gap. The weeks are not on the page yet, so this stays open until they are built and published through the usual gates.
goal Ship an auto-updating weekly work log as honest proof of recent workThe home-project post was checked against the records and its details made generic
A draft post about a drawn-out home project was audited against the underlying records: the permit, inspection and fee details were corrected and made generic, a line that read as leverage over a bad review was dropped, and the fee story was fixed to name a contractor rather than the wrong one. Independent reviewers checked the post against the records before the corrections landed.
goal Keep this site clean, accurate, and currentThree public claims were corrected to match the code
Three claims on the site that no longer matched the code were rewritten to be accurate, covering a project page, a post about tools, and a description of one of my own tools. Each correction was checked against the actual behaviour before it shipped.
goal Keep this site clean, accurate, and currentThe preview tool deletes its temporary site copies once their server stops
The post-preview tool now deletes the throwaway site copies it leaves in the temp folder once their server stops, so previews no longer pile up on disk. The same work cleaned up test servers that earlier deliberate-breakage runs had left running.
goal Keep this site clean, accurate, and current Project: Akaya
10 sessions in this report · active 4/7 days
An AI platform. My work here makes its question-answering measurable and trustworthy: golden datasets and a real evaluation harness instead of guesswork.
during this weekCarried a release through acceptance testing and an upload fix to production, then ran a retrospective on the week's process.
Rebuilt how Akaya's answers draw on source documents
Rebuilt how the question-answering library carries details from source documents into its answers, checked with click-through tests. It works in the practice environment, and the production release is the next step.
Ran an Akaya release
Released a new version of the app, confirming the pre-checks before each step, and followed up on the monitoring afterwards.
Independently reviewed several Akaya pull requests
Several pull requests were reviewed in fresh contexts, with each claim checked against the code and the continuous-integration state before a verdict was posted to the pull request. Where a change needed a small fix before merging, the review said so; the rest were cleared.
Project: Personal
8 sessions in this report · active 4/7 days
The life and money admin that keeps the runway clear while I build.
during this weekWorked through a home insurance renewal and kept the billing record current.
Worked through household financial planning and a contract
Worked through a stretch of personal financial planning: reviewed a home-services contract clause by clause against what had been agreed, compared insurance options as one policy was ending, built a calendar of upcoming payments, chased an outstanding invoice, and cleaned up cloud storage that was near its limit. Each step ended with a clear next action rather than an open question.
Built automatic import of bank and card transactions
Built a way for bank and card transactions to reach a personal finances system automatically, so a monthly sign-in to each bank portal is no longer needed. The provider’s coverage was confirmed for every account, the sync was tested end to end against a demo key, and an independent review landed it after catching a demo key that could have leaked into real data and a gap in backfilling a stale account.
Set up card and bank-transfer payments for the practice
Set up a way for clients who do not come through an existing marketplace to pay the practice by card or bank transfer, confirming the provider’s current signup rules and fees and checking that the services are not taxed in the relevant state. The setup was written up as a document so the remaining account steps are clear.