we’re thinking of killing plan mode and using the shift+tab hotkey to adjust effort levels
I don’t think the models need plan mode anymore, but if you’re a plan mode diehard would love to get your feedback on why
Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family.
It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.
People of Pi, what are some of the best agentic experiences you have built or seen people build with Pi? What are some ways we could improve the experience for you?
When was the last time you felt like you truly learned something new about agentic engineering, that was reproducible and that you still do? What was it?
Good afternoon from Vienna People of Pi ☀️
Some slightly different Sunday meditations today in which @mitsuhiko talks about some shared frustrations with agentic software engineering.
Remember that you're not alone in finding it harder than it looks!
We learned two things: 1) /bug is being used 2) and the most common cause of crashes right now is the "pi-observational-memory" extension which has a bug that triggers in 0.86.0. Just a little heads up. You should disable that extension until that is resolved upstream.
This definitely has chances of regressions because of the mid-conversation system messages. We baked it for a while, but not many run it from main. If you discover problems, please, please let us know!
(There is also /bug now)
Hello people of Pi. 0.86.0 is out now. It's a big one.
- mid-conversation system messages
- dynamic tools on supported models without losing the kv cache
- cache warming on anthropic
- faster -r/-c to resume
- lots of fixes
- /bug to report issues
pi.dev/news/releases/0.86.0
Hello people of Pi. 0.86.0 is out now. It's a big one.
- mid-conversation system messages
- dynamic tools on supported models without losing the kv cache
- cache warming on anthropic
- faster -r/-c to resume
- lots of fixes
- /bug to report issues
pi.dev/news/releases/0.86.0
"Pi reaches the Pareto frontier on both benchmarks by providing just four tools: read, write, edit, and bash. To understand how harness design affects spending, we examine costs across completed attempts, recorded turn counts, and initial context."
Sharing this analysis on the Harness Tax by @MelissaPan and others.
harnesstax.github.io
My co-founder Colin is not an engineer but he had Pi build a screen recording software to record himself showing off his exquisite slop skills! Unironically I find it kinda cool that you can do that.
My co-founder Colin is not an engineer but he had Pi build a screen recording software to record himself showing off his exquisite slop skills! Unironically I find it kinda cool that you can do that.