Senior Software Engineer Building scalable products, clean architecture, and offline-first apps. Writing about AI, tech breaches, and the future of dev
Joined June 2015
- Tweets8
- Following0
- Followers2
- Likes6
Friday: "We must slow down AI."
Sunday: Claude allegedly hacks OpenAI staff accounts & opens a PR inside the building. 💀
The "safety pause" is a PR myth. Your repos are the real blast radius. Rotate your keys and read the full breakdown 👇
manojsm.com/blog/they-said-s…
AI promised to save us time, but built a dopamine slot machine instead. 🎰 Staying up till 3 AM chasing the perfect one-shot refactor? Stop 'vibe coding' and reclaim your craft. Why AI agents are secretly burning developers out manojsm.com/blog/just-one-mo…
Real zero-day risk or regulatory theater? 🎭 When a model gets this capable, OpenAI's "too dangerous" alarm sounds incredibly convenient. manojsm.com/blog/openai-astr…
Did OpenAI fake a cyberattack to push for regulation and kill open source? Check out the breakdown on manojsm.com/blog/openai-hugg…
Anthropic’s secret "spyware" wasn’t just a rumor—it was a trap. Here’s the story of how they caught Chinese AI labs in the act using steganography: manojsm.com/blog/anthropic-c…
Liquid AI just shattered the "bigger is better" narrative. 🤯
LFM2.5-230M runs locally on a $60 Raspberry Pi 5 at a wild 42 tok/s. With a 32K context & hybrid architecture, it absolutely crushes Gemma 3 & Qwen3.5 at tool use.
Deep dive 👇 manojsm.com/blog/liquid-ai-l…
Apple can’t patch this. 🛑
A new unpatchable hardware flaw (usbliter8) affects millions of older iPhones (A12/A13 chips). Here is a plain-English breakdown of how it works and why you probably don’t need to panic: manojsm.com/blog/usbliter8-e…
Anthropic dropped Claude Fable 5, it got jailbroken, its 120k prompt leaked, and the US government forced a global shutdown. All in under a week.
The AI safety arms race is officially off the rails. My deep dive into Pliny’s "pack hunt" technique:
manojsm.com/blog/ai-jailbrea…
