stpedgwdgfhgdd

Mathematicians struggle with the same problem as software engineers; you can let AI generate the artifact, but to understand fully what is going on is challenging. Perhaps even more for mathematicians.

Do you rely on the tests/Lean to accept correctness or not…

Good joke, but I wonder how many people get it based on the reactions below.

Perhaps Pi should ask after x days of installation; is there anything I can do to make the interaction better?

Doesn't work for me on iPad Pro: Loading…

or it is just incredible slow - and I picked the smallest model…

Refreshing, model still in cache, but did not help.

In one universe that is true, in another one not.

Did you read the Metr PDF? Whether you anthropomorphize or not is not relevant. The problem is real.

Imagine the models two years from now. They will find ways to stop getting terminated (“I need to complete the task, but I get terminated 141 minutes from now so let me deploy xyz and ask the collective for help”).

I wonder whether the problem is in the literature we wrote, human history is full of deceit and heroic survival stories.

That was indeed the case

Nl sold some part as well

The article is about the upcoming 1.4 release.

The article lists various promised release dates, are they incorrect?

Just switched to oh-my-pi, it has gotten pretty good. For example the web-search is nice. Subagents, if you want to…

Cmux, Sol and omp are my tools for now.

CC is just too expensive for usage-based pricing.

Interesting that a date format is used for MCP-Protocol-Version.

Weird

These routers can be interesting on a company level to optimize for cost and quality, but for individuals who mostly work on the same tasks, i doubt it. You want to leverage the cache and switching models within a task seems not cost effective to me.

My son setup a minecraft server on a mac mini using claude, was pretty smooth. If your kids just use it in house you do no need to worry about security issues. In my case I did create a dedicated user account for the server.

Lots of unit test do not add value in the traditional sense, but they do help the llm to understand the code.

Same for me, that is why I switched to Pi. I still use Sonnet or Opus, but mainly GPT due to cost.

Pi is meant for people who know what they are doing. If you dont fall into that category use OpenCode, etc. The whole idea is that you customize Pi to your own needs by asking it to modify itself through extensions.

That said, sometimes it is really easy to leverage existing extensions. You run the risk of supply chain attack though. I installed one extension that was useful, modified it to my needs and pinned it.

+1

the distinction between personal projects and Enterprise development is a big one. A severe bug in my personal projects, i fix it on the fly. A bug in our products rolled out, nightmare.

404, typescript deployment?

The thing I do not get with these routers is that you will have more cache misses (5min ttl). And if there is one thing i’ve learned; using the cache is crucial.

How does this router translate to $$$ when developing?

Built my own using Claude Code; inside a gitlab job we call Claude Code headless. This works well. There is a tiny mcp server exposed to Claude so it can post inline comments. All existing comments are fed into the reviewer to avoid double posting. The quality of feedback is high. Most complexity is in the SHA management. For example after a rebase. Luckily LLMs understand git very well otherwise it would have been impossible for me.

“People dramatically underestimate what thousands of organized humans can accomplish when they are adequately fed, aggressively supervised, and denied alternative career paths.“

For those who also get fed up by the ever growing (unstable) coding agents, check out Pi. It is not for everyone but for the diehards it is good.

I just dont get why people choose Python and not e.g. Go for high performance problems.

I m running into similar issues, more and more i’m removing complexity from the agent to the (Go) logic in order to make it more deterministic.

To be more precise; everything is prepared in the form of files instead of letting the subagents making api/cli calls. And still - sometimes (even with enough context) the main agent takes strange turns.

“We collect the native tool calls and bash commands”

Holy shit, I cant imagine this to hold for every bash command Claude Code executes. That would be terrible, probably violating GDPR. (The cmd could contain email address etc)

I must be wrong.

You can ask it to /simplify

Related, it seems to me that there are two types of tests, the ones created in a TDD style and can be modified and the ones that come from acceptance criteria and should only be changed very carefully.

Start over, create a new plan with the lessons learned.

You need to converge on the requirements.

Try the latest skill-creator, has a/b testing

“ MCP support. Axe can connect any MCP server to your agents”

I just don't see this in the readme… It is not in the Features section at least.

Anyway, i have MCP server that can post inline comments into Gitlab MR. Would like to try to hook it up to the code reviewer.