0xsojalsec

Nearly 1/3 of gen Z has already passed away 😐

Both models outperform Opus 5.5 in almost every capability it evaluated, including coding, reasoning and vision.

Power draw: 2-3W (runs off battery, screen off), Debian 13 chroot & Magisk, No extra hardware.

Google is on a run! Google has introduced two new TPU systems: 8tx for confidential AI workloads and 8i Max for large context windows.

DLSS 5 is actually a amazing technology,

Guys Opus 5.5 feels nerfed, incoming model appear in line Fable 5.5 & OpenAi GPT-6.2 Already in route.

The real winners will ship Hardware, OS , Inference stack as one plug-and-play system.
Plug-and-play local inference is about to become a real business.

- Anthropic prompts to master Opus 5.5
- /harness for Managed Agents
- /loops + /self-improving agent teams
- Andrej Karpathy's method
- 24/7 self-improving graphs on max

China just achieved it the world’s first 1,000 km hollow-core fiber-optic breakthrough speeds up to 30% faster than traditional cables.

The higher your GPU tier, the more likely you are to actually shape the future.

Nvidia CEO Jensen Huang had a romantic relationship with Alice Huang they suspect China recruited to spy on him.

In within one Week.

A memo on this reached a few senior officials just before Trump's China summit in May.

a script that emails Sam Altman to let him know when their RAM sits idle.

Sam Altman said the RAM belongs in datacenters, that the RAM in people's PCs is used maybe 5% of the time, and called it quite terrible.

Ram-Altman logs a Linux PC's memory every five minutes and, once a day, reports in if more than 10% of it is unused.

Anthropic engineer Addy Osmani warns that AI could make us stop thinking for ourselves.

Working on Claude Code, he says AI agents now handle so much of his work that he rarely needs to use critical thinking.

He identifies three stages: cognitive offloading (delegating tasks), deferral accepting AI answers and cognitive surrender trusting AI completely).

Skills that you stop exercising are going to go away, and probably faster than you might guess.

This is Stanford CS329A Self-Improving AI Agents (Part 1 / Course Overview), taught by Aakanksha Chowdhery and Azalia Mirhoseini

Azalia is recursive Ai Lab founder, Anyone can build an AI agent in 60 minutes

- Build your first AI agent
- Create agents without coding
- Make $100K+ per month with agents

Claude Code 9 steps, one page, and your agent stops needing you

Paste it into Opus 5.5 tonight and wake up to finished tasks with every test green

99% of people still type ‘fix this bug’ and babysit every step,

Google just dropped a free 2-hour course on Harness, Loop and Graph engineering

From one prompt to 100+ agents that test, judge and fix their own work:

- Harness engineering from scratch
- Build your first agent the right way
- Evals: an agent that grades its own work
- Loop engineering: send bad work back until it passes
- A self-improving graph that gets better while you sleep

Prompts, Agents, Harness, Loops, Self-Improving Graphs

Worth more than any paid agentic engineering course.

A non-public Anthropic model to prove the Aaronson-Ambainis conjecture, a central problem in quantum computing that had remained open since 2008.

The result helps explain when quantum computers can achieve an exponential advantage over classical computers.

PSI says the system found a complete proof on its third day, and the final version was checked with Lean.

Finance jobs are among those most at risk from AI, per Bloomberg:

Ai Speeds Up Ai,

Anthropic’s newest internal model 4 just cleared 50% on every benchmark METR tested.

METR also estimates Anthropic is now progressing 2-3× faster than the historical rate and 80% of that acceleration is coming from AI itself.

New Anthropic model (finished training last month): AI is now the main driver of the speed-up.

Canva is laying off 22% of its workforce which is 1271 employees, and this is what their CEO said.

- give them 1 year of LinkedIn Premium & career coaching. 💀