Developing Dosu github.com/devstein
- Tweets 724
- Following 1,242
- Followers 734
- Likes 491
never tell an agent to "be thorough"
we found a "be thorough" in an internal agent system prompt, removing it decreased spend by 67% with no significant change to quality 🤯
1
1
15
813
one of the biggest challenges of agent knowledge MCPs is showing the impact of it to users
if done right, memory just works and makes your agent better, but its invisible to users.
we've been experimenting with allowing agents to give feedback at the end of their sessions on how @dosu_ai impacted their trajectory: "qualitative agent feedback"
the results are early, but really interesting!
1
6
532
we found the same on our internal benchmarks
for many real-world eng tasks Opus 5 costs ~2x Opus 4.8 for similar performance
that said, I prefer Opus 5's code style
This chart says so much....
- Literally the exact same prompt.
- All long horizon one-shots.
- Totally reflects real-world experience.
2
3
9
1,270
We ran internal benchmarks comparing @OpenAI's GPT 5.6 Sol vs @AnthropicAI Claude Fable 5
GPT 5.6 Sol performs on-par with Fable at a lower cost, but how it does this is fascinating.
GPT 5.6 spends MORE time and tokens gathering context and less on formal planning before executing.
This a stark contrast to previous GPT Codex models that would spend significantly less time gather context compared to Claude models.
It highlights the importance and impact of context gathering for coding agents. Better context, better outputs.
It seems like @OpenAI may have finally found out Claude's secret sauce.
6
13
157
22,822
experimenting with a "Just DM Dosu" short video idea with @OpenAI Sora - wdyt?
1
2
9
573
we got Soham-ed at @dosu_ai
for those that don't know his GitHub profile is
github.com/satya-nutella
PSA: there’s a guy named Soham Parekh (in India) who works at 3-4 startups at the same time. He’s been preying on YC companies and more. Beware.
I fired this guy in his first week and told him to stop lying / scamming people. He hasn’t stopped a year later. No more excuses.
2
6
24
6,111
Thank you @CeleryOrg! We are on 5.5, but we also wanted better DevX for automatically serializing Pydantic models when invoking a task
github.com/celery/celery/iss…
1
3
32
🤌 beautiful new sidebar animation by @caspian_1016 live in app.dosu.dev
2
6
599
ask an LLM to label a PR...and they tell you they are going to a concert tonight?
ALT LLM: I'm so excited to be going to the concert tonight! It's going to be a blast.
2
249
LFG! We’ve been accepted to the #WAGBI developer program🚀 We’ll be using our grant from @alchemyplatform to make it simple for any developer to add web3 auth to their app.
Check us out at picketapi.com/ and follow us @picketapi 👷🛠️
4
19
Built YCRM with @deno_land's Fresh framework and Google PageSpeed loves it ⚡️
ycrm.xyz/
ALT YCRM got a score of 100 on Google PageSpeedInsight using Deno Fresh
1
