progressoverperfection

what if it works out?

[progressoverperfection @nevermindeither]

@ClaudeDevs @claudeai Teams: one reusable workflow installs it org-wide, a fleet dashboard watches every repo's pins. $0 API on a Claude Pro/Max plan. Zero npm dependencies, 98 tests.

One command to start:
https://t.co/lAHpwaEkuo

[progressoverperfection @nevermindeither]

config-drift-checker hit 1.0 this week, and something happened that I did not plan when I started 30 days ago:

the same test suite now runs on @claudeai, @OpenAI 's Codex, and @Google 's @Gemini . Live runs on all three.

Your agent conventions are code. This is their CI.

[progressoverperfection @nevermindeither]

@ClaudeDevs @claudeai It also watches back: every Claude Code release gets a public verdict, subscribable as a feed. The behavioural changelog nobody publishes.

https://t.co/IzIZBajA8y

Current streak: 3 releases clean, about 24 days.

[progressoverperfection @nevermindeither]

@OpenAI @AnthropicAI @Gemini @stripe @Shopify @twilio @SlackHQ @Meta @X @github @gitlab @opentelemetry What went wrong this week: my own CI went red because two records in the deprecation data passed their dates. A tool about dates, tripped by dates.

The data now opens its own pull request when a date passes.

Check your repo: https://t.co/4zi7XbRCIf

[progressoverperfection @nevermindeither]

@OpenAI shuts down 4 models on Monday, September 28: babbage-002, davinci-002, gpt-3.5-turbo-1106 and gpt-3.5-turbo-instruct.

Then 10 more on October 23, gpt-4-turbo among them. @OpenAIDevs #OpenAI

I build DocsWatcher, an open-source check for exactly this.

[progressoverperfection @nevermindeither]

@OpenAI @AnthropicAI @Gemini @stripe @Shopify @twilio @SlackHQ @Meta @X @github @gitlab New for teams: a GitHub App with a dashboard, email or Slack alerts 30 and 7 days before a date, and a view of which dead calls production actually makes, from the @opentelemetry you already run.

[progressoverperfection @nevermindeither]

@OpenAIDevs If you still run gpt-3.5-turbo-instruct somewhere, what is it doing? I want to know what kept it around this long.

[progressoverperfection @nevermindeither]

OpenAI shuts down 4 models on September 28 2026. That is in 5 days.

babbage-002
davinci-002
gpt-3.5-turbo-1106
gpt-3.5-turbo-instruct

After that, requests to them fail. #OpenAI

[progressoverperfection @nevermindeither]

@OpenAIDevs To check a repo, paste a public GitHub URL into DocsWatcher. It's my project. The scan runs in your browser in WebAssembly, so no code goes to a server.

https://t.co/4zi7XbRCIf

[progressoverperfection @nevermindeither]

@ClaudeDevs @claudeai Start in one command:

claude plugin marketplace add jameskomo/config-drift-checker && claude plugin install config-drift-checker@jameskomo && claude "/config-drift-checker:setup"

Teams: one reusable workflow rolls it out org-wide, and a fleet dashboard watches every repo's pins

[progressoverperfection @nevermindeither]

Your @claudeai setup is code: CLAUDE.md, skills, hooks & it breaks silently, because @claudeai ships new releases a month & the models change.

config-drift-checker is CI for that setup: tested on every PR and every release, against a pinned baseline.
#ClaudeCode #BuildInPublic

[progressoverperfection @nevermindeither]

@ClaudeDevs Built on @claudeai official claude plugin eval, not beside it: same case format, official runner underneath. We add what it leaves out: a stored baseline, history across releases, noise bands that tell flakes from breaks, a canary per release, pin bump PRs, hard budget caps.