latest
Claude Code was using 51,000 tokens before I even typed a prompt â I fixed it
My fresh sessions weren't actually fresh
I put a local model in charge of naming and filing every download, and my Downloads folder has been empty for a month
No more yearly spring clean to get rid of the junk
My local AI model was wasting VRAM by default, and changing one setting doubled its speed
When it comes to local AI, context matters for more than one reason
I hooked up my AI browser to a local LLM and it finally solved my biggest problem
When you stop letting your browser AI phone home
I use Claude Code for planning but a local model for building, and I never hit my usage limits anymore
My usage limits finally stopped being a problem
Claude Code kept shipping broken code until I added this one automation
I just needed to add one hook.
I gave Claude Opus 5, GPT-5.6, and Grok 4.6 the same complex web project and only one behaved like a senior developer
One planned ahead, two just coded.
I gave the same bug to Claude Code, Codex, Antigravity, and their open-source rival â only one handled it like a pro
Reliability is not a guarantee
I used Claude Code to improve my operating system, and here's how it made me more efficient
Coding was only the beginning
I built a fully local agentic AI setup with one open-source app, and it never touches the cloud
The agent setup I trust with folders I'd never hand to Claude
I turned two mini PCs into a local AI that reads my whole codebase, and now it reviews my code for me
This local LLM automatically opens GitHub issue reports for me.
Open-source maintainers are drowning in AI-generated garbage, and it's getting worse
Creating is easier than checking code over.
5 features Claude used to make you pay for that are now free
Yesterday's premium features are today's freebies
My phone runs an AI assistant entirely offline, and I stopped uploading my documents to cloud chatbots
I can finally trust mobile AI with the important stuff
I ran Claude Code against Archon on the same bug six times, and the run that failed told me the most
The fix was never the problem
I stopped dumping raw sources into NotebookLM, and it got so much better
The source was the problem all along
I found this Claude feature so good that I canceled ChatGPT, Gemini, and Perplexity
The AI feature I actually needed.
Unsloth's new desktop app does the one thing LM Studio and Ollama never let me do
It lets me fine-tune the models.
Auto mode is now Claude Code's default, and it changes how fast you can actually build
And surprisingly, it actually beefed up security
