I really like the term “context engineering” over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM.

Claude Code pro tip: Ask it to use Gemini CLI with its 1M context window and free plan in non-interactive mode to research your codebase, find bugs, and build a plan for Claude to action Prompt and link below:

Quite an interesting social experiment by Anthropic in a real world setting using Claude Sonnet 3.7 as an autonomous agent named 𝘊𝘭𝘢𝘶𝘥𝘪𝘶𝘴! anthropic.com/research/proje…

Small Language Models are the Future of Agentic AI Lots to gain from building agentic systems with small language models. Capabilities are increasing rapidly! AI devs should be exploring SLMs. Here are my notes:

When optimizing a loop in Python, a performance trick is to create a local alias for global functions or variables used in the loop E.g.: l = len for _ in range(1_000_000): l(lst) But we still need it given so many perf improvements in Python? I did a benchmark and it appears

This seems pretty cool, gotta test this experience against Coolify!

I blogged about systemd because I finally snapped after reading Reddit threads about it one too many times "systemd has been a complete, utter, unmitigated success" ⬇️

bitchat now on android. open source moves fast.

Gigachad!

After 2 hours of debugging, now I know why the hell fan of my ICE Tower Cooler is not working when I'm booting through M.2 SATA SSD yet working via SD Card boot.🫠 SSD causing momentarily voltage drop across the USB to which CH340 chip is also connected. (but why, gotta figure)

Someone made a cool video about the AI 2027 report, though I don't align with the timeline but this could possibly be how the events will largely unfold. One analogy related to Agent-3 in the report - "Boeing after merging with McDonnell Douglas😁" youtu.be/5KVDDfAkRgc?si…

Couldn't put it better! 🫡

Great results, definitely on my radar!

this screenshot from GPT-5 livestream has to be among the worst chart crimes of the century

Totally agree, reasons I'm majorly scouting the undervalued markets like China, Vietnam and maybe Thailand as well.

RL to rescue, as Always!

To run gpt-oss-20b on a 16GB Mac use these commands: brew install llama.cpp llama-server -hf ggml-org/gpt-oss-20b-GGUF --n-cpu-moe 12 -fa -c 32768 --jinja --no-mmap Then open the browser at http://127.0.0.1:8080

Just been cleaning my GitHub notifications before starting my new job next week (eee!) and figured I haven't mentioned my silly little `spacer` CLI tool in a while. It adds spacers when command output stops so you don't have to hit enter a bunch of times. github.com/samwho/spacer