Head of OSS @dosu_ai
Joined November 2008
- Tweets 10,957
- Following 1,599
- Followers 3,572
- Likes 45,898
2,942 Photos and videos
Jun 11
Jun 11
Weāve reached an agreement to acquire @ona_hq.
Its secure cloud execution technology will help Codex take on longer-running work, even when laptops are closed, and help more organizations deploy agents securely in production.
After closing, Ona will join OpenAIās Codex team. openai.com/index/openai-to-aā¦
1
4
326
Jun 5
A single project is a roadmap you trust someone to follow, while an ecosystem is a hedge. AI-native networking needs an ecosystem. linkedin.com/pulse/why-ai-naā¦
1
1
64
Jun 4
AGENTS.md is the new README for AI coding agents. Over 60K projects use it, and despite the ecosystem being a bit messy, there's still a lot to consider as you write instructions for agents. bit.ly/4pPbEgf
3
178
Jun 4
When every candidate has an agent, the whiteboard interviews don't work like they used to. It's all about judgment and taste now! steve-yegge.medium.com/the-lā¦
87
elixir has a type system and gta6 isnt out yet
Jun 3
Elixir v1.20 released! Now officially a gradually typed language: Elixir type checks every single line of code, finding bugs and dead code, without developer overhead (no typing signatures) and extremely low false positives rate. Plus a faster compiler! Links and reports below.
2
12
116
8,032
Jun 3
Most AI policy is written from a security, compliance, or cost perspective. Though @skamille is approaching this by respecting people's time (a far better approach IMHO) skamille.medium.com/guidelinā¦
46
Jun 3
Cutting Claude Code token use by ~90% is WILD, but check out those tradeoffs. Cheaper context isn't "free" when you pay for it with recall. (Or, you can just use @dosu_ai š) getpushtoprod.substack.com/pā¦
104
Jun 2
Your agent isn't dumb. When you throw it a whole library and ask it to read every book, thatās likely hurting your outputs. Most AI problems are bad context problems (weāre deeply passionate about this at @dosu_ai).
medium.com/@Gal-dahan/your-aā¦
43
Jun 1
When you make coding agents think wide before deep. Or branch first, prune second. Then you can prevent issues, and find more novel paths than before! Diverge then converge, my friends. uditakhourii.github.io/adhd/
1
1
92
Jun 1
AI can help translate between business, product, and engineering. But itās unable to handle the nuance of negotiating between all those teams.
ajeygore.in/content/the-anatā¦
ALT AI figuring out the anatomy of an AI native org.
78
May 31
"A stale AGENTS.md can be worse than no file at all."
- @cloudflare
Missing instructions cause your agents to ask questions. But poor instructions make your agents act confidently and incorrectly. How can we fix this?
dosu.dev/blog/a-stale-agentsā¦
1
2
183
Taylor Dolezal retweeted
May 30
The more I use AI, the less I feel itās just about speed. Itās more like, you can take your own work a little further than you normally would. Small ideas suddenly get room to become way more interesting. @dosu_ai
2
8
452
May 30
Multi-agent failure mode! Not only a great band name, but worth thinking about how orchestration is mostly about explicit handoff contracts.
grammarly.com/blog/ai/types-ā¦
50
Taylor Dolezal retweeted
May 29
open models will become a huge force for coding.
May 29
The latest finding in the LangSmith Signal: Open Models are having a moment.
1 in 3 AI teams ran an open-weights model in April 2026, up from 1 in 5 nine months ago.
The overall number of teams using open weights grew 3x.
Weāre seeing newer users choose open models at a higher rate than those who came before.
8
14
68
19,730
May 29
Token cost is easy to measure, but answer quality is not, so if you reward solely on cost or token reduction, output quality can slip for a while before it's noticeable.
Keeping answer quality high while cost comes down is most of what we work on at @dosu_ai. Happy to swap notes on how to measure the quality side!
May 29
Yesterday I wrote how I'd expect more engineers to get rewarded the next perf reviews and promo cycles for saving token costs for their company.
Today already got messages from devs who already got rewarded with eg bonuses for having done exactly this
It only makes sense!
1
149
May 29
āWhat happens when every company has access to the same model? The best riders win.ā
tomtunguz.com/harnessing-ai/
1
70
May 28
Great post from @digitalocean on prompt caching to help avoid reprocessing expensive AI requests for the same context.
digitalocean.com/blog/promptā¦
1
203
May 28
At @dosu_ai, we solve that problem at the knowledge layer by capturing shared, up-to-date context from code, docs, Slack, GitHub, and more.
That allows you to focus on your work, while we handle the knowledge infrastructure.
169