Going to start posting a bunch more stuff you can do with agents. Basically spent the last 12months+ working with and building agents and tools for agents. Will start sharing all my shit work. Dont fight me, i'm just talking to clankers and they seem to talk back alot.
Everyone talking workflows but i've been using pi to launch as many subagents or agents as i want, with any model or any provider I want since last year. Why is everyone acting like claude workflows are the coolest thing. Its old shit. Mario and nico gave me this for free.
Thought I needed at least a 3090 to do this but following @0xSero helped. With enough context engineering and a good agent, you can simply do things even on 8GB. Now hunting for 3090s on facebook marketplace.
Pushing the limits of my local LLM on 8GB VRAM. All experimental shit, NFA. I also pay for Claude and chatgpt max plans but i am slowly weaning off. Qwen 3.5 9B is definitely not close and my measly 8GB VRAM is definitely not enough but we can do light tasks.
Eli Lilly released retatrutide Phase 3 data yesterday. 28% weight loss in 80 weeks. The most powerful obesity drug that’s ever been tested.
And today the cancer signal drops.
12,112 patients. Seven tumor types. GLP-1 users had half the lung cancer metastasis rate (10% vs 22%). Breast cancer: 43% cut. Colon cancer five-year mortality in a separate study: 15.5% vs 37.1%.
Cancer joins a list that already includes heart disease (SELECT, 20% MACE reduction), kidney failure (FLOW, 24% slower decline), sleep apnea (SURMOUNT-OSA, FDA-approved), addiction (BMJ, 600K veterans, 18-25% reduction across substances), and liver disease (86% fat clearance).
Tumors express GLP-1 receptors. Activate them and NF-kB drops, apoptosis rises. The drug isn’t just shrinking fat. It’s talking directly to the cancer.
One drug class. Designed for blood sugar. The biology keeps finding uses the designers didn’t predict.
I have only written software and integrated systems the last decade of my life. So anytime i see a breadboard, i have fond memories of my engineering days and radioshack.
I love Pi. So I created a memory upgrade for Pi coding agent, Pi already has trees & session logs with this complementing it. Including an auto-compact feature to help you manage your context. There is probably one out there already but this one is mine. https://t.co/sL6lGqkJd5
My smol take, EVs existed in the 1900s but didn't quite pass the test of time, fast forward to 2010, we have tesla and EVs here long term. Sure uchat wasn't maintainable but they didn't have AI models. EV didnt last because we didn't have superchargers. Claude = Supercharger.
Cannot wait for teams that build their custom JIRA/Workday replacement/custom CRM to one day turn around and ask:
“Why do we have so much internal software that is buggy / has poor UX + we need to maitain?”
Seen this movie well before AI, when Uber built uChat (custom Slack)
Perplexity computer is quite depressing because i know Claude or openai will make a better model in a week.
At this point, if i'm perplexity whats the point.
Aside from being an impressive coding agent. I'm using Pi agent for all sorts of usecase. A simple one was using to as an interactive teacher for karpathy tinyllm. Combed through section by section, compare with gpt 2 & 3. Cool.
Love the fact that Chinese labs are doing good/fraudulent work but i don't care. All in the name of progress and competition. However, i have used https://t.co/c8uQ2nWYO8 and Deepseek and I have to say, it's nothing close to the quality of OAI and Anthropic. Distill all you want
Trying to move microfocus cobol to cobol as unix version approaches EOL. No one knows who wrote this shit that somehow the whole organization uses for a matching engine. Its sitting in 25years later, doing exactly what it was intended to do. Now its my headache to modernize.