The #environmentalimpact of #AIcodingagents like #ClaudeCode is significantly higher than typical LLM queries. A single Claude Code session involves numerous tool calls and long prompts, leading to substantially more energy consumption. The author estimates the energy cost of a Claude Code session by analysing token usage and applying pricing data from Anthropic’s API. https://www.simonpcouch.com/blog/2026-01-20-cc-impact/?eicker.news #tech #media #news

This #guide provides a #framework for writing effective #specifications for #AIcodingagents. It emphasises starting with a high-level vision and letting the AI draft the details, structuring the spec like a professional PRD, and integrating the spec into the development toolchain. The guide also highlights the importance of planning, breaking tasks into smaller chunks, and continuously iterating on the spec and implementation. https://addyo.substack.com/p/how-to-write-a-good-spec-for-ai-agents?AIagents.at #AIagent #AI #ML #NLP #LLM #GenAI

Interesting read https://www.the-main-thread.com/p/ibm-bob-ai-partner-enterprise-java-codebases #aicodingagents

Michelle Bakels (@MichelleBakels)

AI 코딩 에이전트 구축 관련 강연 요약과 코드 평가 현황을 다루는 내용입니다. 'Hard Won Lessons from Building Effective AI Coding Agents'(Nik Pash, Cline) 영상과 'State of Code Evals'에서 SWE-bench, Code Clash 및 최신 코딩 벤치마크(SOTA)들을 정리한 John Yang의 요약이 포함된 링크를 공유하고 있습니다.

https://x.com/MichelleBakels/status/2008934971495297359

#aicodingagents #codeevals #benchmarks #swebench

I've been trying to make sense of staff/chief/distinguished level engineers being enthusiastic about #AICodingAgents. Intuitively you'd assume they're the adults in the room and should easily detect the quality and bullshit that such agents pull.

Mindsets I can explain:

Junior engineers: oh my, I'm getting super powers!

Mid engineers: I thought I had so much left to learn, but maybe I don't have to!

Senior engineers: this looks cool, but hang on let's have a closer look 🤔 — err... wtf!?

Staff engineers: now I no longer have to work with junior-senior engineers to work out the details!

That last bit is key, I suspect. At the Staff level, you're less involved with the details, mostly with the bigger picture and bigger ideas. At that level, AI code looks fine. You're used to having people you can leave ironing the details to, now those are just automated. The trap here is that the people you handed work to weren't professional bullshitters. Likely you had trained them yourself, you could gauge their skillset and knowledge or get hints of what they were based on their communication. That's no longer the case with coding agents. They always sound "PhD level" no matter what. Therefore it's very easy to miss the garbage they produce.

In some contexts I acted at a Staff-level and it took be considerable time to "poke" through this reality distortion. Maybe all those enthusiastic Staffers we hear now are simply not there yet.

The sooner the beter, though. I'm still cleaning up the mess after my own "AI coding power trip" in the @silverbulletmd code base.

https://mas.to/@carnage4life/115832534415373032

Vibe Kanban – Kanban board to manage your AI coding agents

https://github.com/BloopAI/vibe-kanban

#HackerNews #VibeKanban #KanbanBoard #AICodingAgents #Productivity #Tools #GitHub