It's Dosu Drop Day! 🎒 For our back-to-school edition, Dosu does the studying. When a Claude Code, Codex, or Cursor session ends, Dosu reads it and writes down what your team would otherwise rediscover. 🧵 go.dosu.dev/drop-sep26-x
2
3
6
349
The team keeps working after the agent’s session ends.
@devstein64 on how merged PRs and resolved conversations keep agent memory current.
Recording + transcript: dosu.dev/webinars/agent-memo…
1
2
12
201
Moving the cursor position to the end of Copilot's inline-suggestion prompt helps GitHub reuse more of the cache. I like this example because we can reason about it by looking at what changes between requests.
code.visualstudio.com/blogs/…
ALT GitHub article graphic showing three separate inline suggestion models replaced by one unified model in a code diff.
2
133
Taylor Dolezal retweeted
I'll say this:
I've gotten more out of using orbs and asking agents
"test this e2e & give me irrefutable proof it works"
than out of 80% of tests I've written and ran over the last 15 years.
It's truly remarkable that we can now ask machines:
"Did you manually test it?"
I agree with @thorstenball, I think unit tests are dead in the water.
The ones the models write are terrible, at best just doubling total LOC.
Inverting the testing approach - heavy e2e/black box/golden master, reaching for lower levels only if necessary, works better for me
40
8
357
30,828
Astra, Sol, and Luna from @OpenAI have new badges in Decant v0.6.0. ✨🔥🌙
See where your tokens went and what your agents did in each session, right on your machine.
Your move, @AnthropicAI! A themed Claude lineup would be fun to design around.
github.com/dosu-ai/decant
5
418
ByteByteGo explains why a similarity score of 0.85 isn't 85% confidence. I'd find that easy to misread in a product. Explaining why a result is relevant would help me more than the score alone.
blog.bytebytego.com/p/how-ll…
ALT ByteByteGo retrieval flow. Documents become chunks and embeddings in a vector store. A question is embedded in the same space, matched to passages, and sent to a model that answers with a source.
2
114
Olaf Alders points out that .git is a file in a linked worktree. If a setup script worked in my usual checkout but failed in an agent's worktree, I'd want to see whether it expected a directory.
olafalders.com/2026/09/16/gi…
113
Who reviews agent memory when people can’t keep up?
Michael Mangus, Head of Product at Dosu, on trusting AI to maintain memory—and making mistakes easy to correct.
Recording + transcript: dosu.dev/webinars/agent-memo…
1
3
112
Taylor Dolezal retweeted
it is true
Is it just me or are existing agent memory datasets underwhelming?
They to some extend assume retrieval will just work, they tell you roughly where the right answer is, and then they sit back and see if you can piece together a response.
This doesn't resemble what the work feels like in my experience...
1
3
414
A failure message that shows the unexpected record would give me somewhere to start debugging. Will Keleher's example is a useful reminder that the message itself can make a test easier to fix.
will-keleher.com/posts/its-b…
1
1
160
A summary of a summary keeps getting further from the source.
Michael Mangus, Head of Product at Dosu, explains how repeated summaries lose details in agent memory.
Recording + transcript: dosu.dev/webinars/agent-memo…
1
2
5
123
Alexander Belanger's Postgres guide explains how adding an index can stop writes. I'd want that context beside a migration diff, especially when the small change makes it tempting to skim.
hatchet.run/blog/postgres-su…
1
114
Taylor Dolezal retweeted
Everyone should copy "The Amp Way" for bugfixes:
1. Make it easy to report a bug in-product
2. Fix bugs fast
3. Find every other affected user
4. Email proactive, personalized apologies & credits
5. Get delightful replies & increase chances they report bugs -> virtuous cycle
Try this with one bug fix this week and see how it goes.
20
18
272
25,339
Two agents start from the same commit, make different changes, and both report success. Which version should advance the branch? That's one question I'm exploring in Git with the times, my new Substack piece.
onlydole.substack.com/p/git-…
2
3
376