Looking for open source volunteers anywhere in the world.

Contact me.

Simplex

https://smp16.simplex.im/a#iZux_FXezkIBBNAsS75Lu-6AAzjKRYBjIKE3Gitgb9s

Please use the reference "open source volunteers" in your initial simplex message.

User

#WildFlower

SpaceX on Tuesday announced it entered a formal agreement to buy the artificial intelligence startup Cursor for $60 billion worth of stock, a hotly anticipated deal.

[...]

Cursor built a popular AI coding tool that helps software developers generate, edit and review code, and the company has experienced explosive growth since its founding in 2022.

I am currently super satisfied with Kimi K2.6 and Minimax M3, the quality is great, pricing is low and usage is enormous.

I am however always looking for better deals. I am a power user and use billions of tokens a month. The agents run almost 24/7, sometimes more than one.

What do you use for agentic coding?

I use Kimi K2.6 and Minimax 3. Both are mainly due to pricing.

I got 6 months of Google Pro from buying my new phone, so I use Gemini Pro 3.1 and Gemini 3 Flash quite a bit, but don't think it does too well compared to Kimi and Minimax.

I use Opus 4.6 as I get some usage from Google Pro, but it's just like 1 task or something.

It is obviously the best, but considering the usage is extremely expensive, and I can use Kimi and Minimax almost 100x as much with a subscription. It kinda becomes a worse model. I'd rather Kimi working really long on a problem, than a slightly more powerful model with almost no usage. I can get a lot more done with high quality results from Kimi due to this.

What models are you using, and why?

(Sorry of discussions are not allowed in this community)

The shocking story of Bun's sudden, unexpected Rust rewrite. How it happened, what their strategy is, and the new doors I think this opens for all software developers.

"I wanted to know whether Gemma 4 could replace a cloud model for my day-to-day agentic coding."

cross-posted from: https://mander.xyz/post/50244557

The benchmark is a set of handcrafted 2d puzzle games that are easy to solve by humans, but require features like skill acquisition and long-term planning by agents.___

Dafny is a good intermediate step for LLM generated code.

this is the abstract of the paper:

Using large language models (LLMs) to generate source code from natural language prompts is a popular and promising idea with a wide range of applications. One of its limitations is that the generated code can be faulty at times, often in a subtle way, despite being presented to the user as correct. In this paper, we explore ways in which formal methods can assist with increasing the quality of code generated by an LLM. Instead of emitting code in a target language directly, we propose that the user guides the LLM to first generate an opaque intermediate representation, in the verification-aware language Dafny, that can be automatically validated for correctness against agreed on specifications. The correct Dafny program is then compiled to the target language and returned to the user. All user-system interactions throughout the procedure occur via natural language; Dafny code is never exposed. We describe our current prototype and report on its performance on the HumanEval Python code generation benchmarks.

Some issues can be prevented when vibe-coding, but LLMs find a way of messing up anyway.