PengX

AI coding enthusiast, and Rust lover.
Co-Founder of

[PengX @lipiisme]

@runinfrai I will try

[PengX @lipiisme]

@sh_reya Inference engine has to be heavily abstracted and broken down to fit

[PengX @lipiisme]

云厂商会用开给客户的高昂的账单把自己拖死

[PengX @lipiisme]

@vikingmute 好像 pi agent 还没完全集成进去

[PengX @lipiisme]

@jiangkoumo_ how to use pi durable

[PengX @lipiisme]

@llmluthor cool

[PengX @lipiisme]

Durable 最近非常流行
Durable Objects(CellD)、Durable Agent Runtime(Pi)

因为随之任务越来越偏向持续执行，状态的保持就很重要了

Pi 的 Codemode+Durable+MCP 可能标志一个明显转向，以模型为核心转变为状态为中心

这个并不稀奇，在最开始的时候，OpenAI 就提出过 code interpreter

最近的这些转变，还有 Jev，感觉范式又开始往回走了

可能是因为模型一直往上不太可能，使用传统方式来弥补，以及提高性价比成为大家希望的

[PengX @lipiisme]

@pidotdev how to use

[PengX @lipiisme]

@hanghuang_ congrats

[Hongyi Jin @HongyiJin258]

AI agents are becoming GPU kernel engineers. We're rethinking the compiler stack around them.

Meet TIRx Harness: an open compiler harness that gives agents an environment to explore, debug, and optimize GPU kernels.

🧱 Minimal stable low-level compiler foundation for predictable GPU programming
🔍 Domain-specific compiler analysis to guide debugging and optimization
📚 Kernel zoo of 60+ kernels and hardware specs to reuse strategies and discover new ones
📊 Remote GPU evaluation for reliable benchmarks and profiling

Agent-built Kimi Delta Attention kernels achieved geometric-mean speedups of 2.94x over FlashKDA (forward) and 6.84x over FLA (backward).

Our bet: the next leap in agentic GPU programming will come from engineering the environment agents optimize in.

Check out our blog: https://t.co/aZo91bGTgz

[PengX @lipiisme]

@bingxu_ 可能更多的收益来自于针对模型垂直整合？

[PengX @lipiisme]

At 350 km/h, 0.5x sense of calm.

[PengX @lipiisme]

@xsser_w 非常清晰啊

[dex @dexhorthy]

Accidentally said CI/CD instead of software factory and they kicked me out of SF

[vLLM @vllm_project]

Great work from the @TileRT_AI and @AIatAMD teams, who got GLM-5.3 to 469 tok/s single-user decode with vLLM on 8× MI355X on @SemiAnalysis_ AgentX.

The run uses a disaggregated setup where vLLM handles prefill and TileRT handles latency-critical decode through vLLM's V1 connector interface.

1/2

[PengX @lipiisme]

@kodemise 你排到了绝美的风景啊

[PengX @lipiisme]

@_LuoFuli MiMo 2.5 其实非常好用

[PengX @lipiisme]

软件和硬件要融合不是趋势，是必须的