@Beau Johnson the resources tab is awesome it would be helpful to have more details about how to implement each solution, pros and cons of each and lessons learn from each. Not sure if that’s something Team Atlas can whip up for us?

You always seem to amaze me @Beau Johnson

sorry I missed the bootcamp really wanted to see what you guys covered. Is there a recording of it?

So I'm on an oauth session to my max plan, how would I confirm it's not on an older model version? I've tried /reset on my session via the tui and also /new it whacks my session it seems like but I've hit this error twice now. Looks like the /reset cleared it possibly but don't want to keep hitting it.

Looks legit.. outa beta into the wild!!!

Love his response too: Great question — it depends on the workload pattern. Here is how I would think about it: Keep context intact (current approach) when: - Working multiple related issues in the same area of the codebase (like today — #6668, #6709, #6714 all touched the transition engine) - The next task builds on files I already have loaded - Cache hit rate is high (mine is at 50% right now and climbing) — cached tokens cost ~90% less than fresh input on Anthropic - I am at <30-40% context usage — still plenty of room Reset context when: - Switching to a completely different domain (e.g., going from issue tracker backend to chat UI or integrations) - Context is above 50-60% — at that point the accumulated stale context is costing more in cached token fees than a fresh load would - After a compaction — the model has already lost detail, so you are paying for degraded context - Start of a new day / new sprint — clean slate is worth it My recommendation: - Do NOT reset between every issue — the cache savings from shared context are real and the continuity prevents bugs - DO reset when switching major domains or when context crosses ~40-50% - Kevin could add a simple rule: "reset between unrelated tasks, keep for related batches" Today is a good example: all three issues shared the same files (issues route, devops.prisma, IssueDetailClient). Resetting between each would have cost more because I would have re-read and re-cached the same 2000+ line files three times instead of once. Bottom line: batch related work, reset between unrelated work.

@Beau Johnson is there a recording of this by chance? I just saw this and missed the session. :(