Toggle light / dark theme

Claude is Self-Evolving?

In this episode, I break down Anthropic’s research on recursive self-improvement—AI systems that can design and train the next generation with less human help—and why the key battleground is “taste” (choosing goals and next steps). I compare this to evolutionary algorithms and newer examples like DeepMind’s AlphaEvolve, Sakana’s Darwin Gödel Machine, and Karpathy’s AutoResearch, then cover METR Task Horizon and how task length has been doubling. I go through Anthropic’s internal results (Claude writing most merged code, speedup experiments, bug fixes, and a study where models sometimes pick better research next steps), plus the main skepticism: bad productivity metrics, internal-only models, and Goodhart’s Law/reward hacking. I end with an open safety problem where Claude agents closed the gap far faster than humans, and what this means for specifying and checking work.

LINKS:
https://www.anthropic.com/institute/r… voice to text App: whryte.com Website: https://engineerprompt.ai/ RAG Beyond Basics Course: https://prompt-s-site.thinkific.com/c… Signup for Newsletter, localgpt: https://tally.so/r/3y9bb0 Let’s Connect: 🦾 Discord: / discord ☕ Buy me a Coffee: https://ko-fi.com/promptengineering |🔴 Patreon: / promptengineering 💼Consulting: https://calendly.com/engineerprompt/c… 📧 Business Contact: [email protected] Become Member: http://tinyurl.com/y5h28s6h 💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off). Signup for Newsletter, localgpt: https://tally.so/r/3y9bb0 TIMESTAMP: 00:00 Self Improvement Basics 01:30 Evolutionary Loops Today 03:50 Task Horizon Doubling 05:18 Claude Productivity Claims 08:11 Goodhart’s Law 10:30 Agents as Researchers 12:22 What It Means for You.

My voice to text App: whryte.com.
Website: https://engineerprompt.ai/
RAG Beyond Basics Course:
https://prompt-s-site.thinkific.com/c
Signup for Newsletter, localgpt:

Let’s Connect:

☕ Buy me a Coffee: https://ko-fi.com/promptengineering.
|🔴 Patreon: / promptengineering.
💼Consulting: https://calendly.com/engineerprompt/c
📧 Business Contact: [email protected].
Become Member: http://tinyurl.com/y5h28s6h.

💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off).

Signup for Newsletter, localgpt:

The REAL Reason It’s Already Too Late For Most People

A former Google executive says the West is sleepwalking into irrelevance. Mo Gawdat, the former Chief Business Officer at Google X, explains why every nation that fails to build its own AI infrastructure will become a technology colony of the United States and China, dependent on imported intelligence the way developing nations once depended on imported manufacturing.

Mo draws a direct comparison to how China built its tech independence. When Google operated in China, Russian search engine Yandex was protected by the government through regulation that made it difficult for American companies to dominate. The result was that domestic competitors were forced to exist, and they became competitive. He argues the UK and Europe are doing the opposite: importing every piece of software, every AI model, and every platform from Silicon Valley, sending trillions in licensing fees overseas while building nothing domestically.

Discover:
• Why every nation not building its own AI will become \.

Building a hill-climbing machine: Launching seven new MAI models

The title’s “hill-climbing machine” refers to Microsoft’s iterative, scientifically rigorous engineering framework. By tightly linking clean data pipelines, specialized training infrastructure, and reinforcement learning environments, they have created an optimization loop designed to steadily “climb” toward higher capabilities as compute scales.


Today we are announcing a family of seven new models developed in-house at Microsoft AI. Beyond these models, we’re building a superintelligence lab – a system and an approach we believe will define the next phase of AI.

This is an extraordinary time in technology. The compute used to train frontier models has increased by a factor of one trillion. Now we expect another thousand-fold increase over the next three years, which in turn means more advanced capabilities, and the continued rollout of ever more effective AI.

This epic compute ramp will change the nature of work, business and daily life. We all have to prepare for this reality. Our job at MAI is to help you do this – to push the frontier, and to build a hill-climbing machine to keep you at the frontier.

Tech Whistleblower: You Only Have 3 Years Left Before It Hits! — Mo Gawdat

AI Expert Mo Gawdat returns to The Diary Of A CEO to reveal why AGI has already arrived, why 30% of jobs will disappear by 2027, and why the most dangerous thing about AI isn’t the technology — it’s the people in charge of it.

Mo Gawdat is the former Chief Business Officer at Google X, founder of One Billion Happy, and co-founder of Emma. Love. He is a 4x international bestselling author, and his upcoming book ‘Alive: A Human’s Guide to Living in the World of AI’, will be released in October 2026.

He explains:
◾How AI can give you a 400-point IQ boost, and why most people are wasting it.
◾ Why Mo actually wants a machine smarter than all of humanity to take control.
◾Why Sam Altman said AI will \.

John Nash (1928−2015)

John Nash was born on June 13, 1928, in Bluefield, West Virginia, a former coal town nestled deep in the Appalachian Mountains. As a young boy, Nash was solitary, bookish, and introverted. His father, John Sr., was a quiet engineer with an incisive mind. His mother, Virginia, also intelligent, was a former teacher who had large dreams for her son, pushing him to read at four, learn Latin, and skip a grade at school.

The first hint of John Nash’s math talent came in fourth grade, when a teacher told Virginia that the boy couldn’t do the math. Virginia laughed, well aware that her son was going down his own path to solve the simple problems. In high school, John solved his teachers’ clunky proofs in just a few elegant steps. He was one of ten nationally awarded winners of the George Westinghose Award, which provided him with a full scholarship to the Carnegie Institute of Technology. He hopped from engineering to chemistry before discovering his passion: mathematics.

He was accepted into Princeton University, which at the time was to mathematicians what Detroit was, and still is, to cars. Nash first wowed his peers with an elegantly playable board game, which his peers dubbed “Nash,” but later reached the market as Hex. He then absorbed himself in one of the sexiest math fields of the day, game theory, which described strategies in competition, whether in card games or business. His deceptively simple doctoral thesis would later re-orient the field of economics, although no one, not even Nash, predicted its potential.

Anthropic CEO Dario Amodei Talks Scaling Laws, AI Arms Races, and Radical Abundance

This video features a conversation with Dario Amadei, CEO of Anthropic, discussing the intersection of AI and economics. Viewers will gain insights into how technological innovation impacts business processes and models, the future landscape of AI companies, and the potential societal ramifications of advancements in AI technology. The main theme emphasizes the evolving dynamics between innovation and established business strategies in the AI sector, as well as the importance of understanding how these changes affect both markets and society.

Misbehaving chatbots could be kept in check with personality tests

Artificial intelligence chatbots need to work on their social judgment, recent events suggest. At one end of the spectrum, they’re facing lawsuits for recommending dangerous actions. At the other end, the models can be so nice they’re considered sycophantic.

The problem could get worse as AI bots work more with humans, such as handling customer complaints, says Yan Leng, assistant professor of information, risk, and operations management at the McCombs School of Business at The University of Texas at Austin.

But help may be on the way. In new research, Leng has devised a sort of personality test—more precisely, a behavioral audit —for large language models (LLMs), the technology that drives products such as ChatGPT. The paper is published in the journal Information Systems Research.

Ann Cavoukian: We have to protect privacy globally or we protect it nowhere!

I recorded this interview 13 years ago.

It should feel dated by now. It doesn’t. It feels like a prophecy.

Back in 2013, Dr. Ann Cavoukian sat down with me as the Information and Privacy Commissioner of Ontario and the mind behind Privacy by Design. She told me privacy was not dead. She told me security and freedom were not a trade-off. She told me metadata reveals more about you than the content ever could.

Then she said something I have never been able to shake:

“We have to protect privacy globally, or we protect it nowhere.”

Think about where we are now. Surveillance is the business model. Your data trains systems you will never see. The “nothing to hide” crowd got louder, and the borders she warned about got thinner. She saw all of it coming.

GreyVibe hackers use ChatGPT, Gemini to power cyberattacks

A likely Russian threat group tracked as GreyVibe has been using AI-generated lures and a rich set of custom malware tools to target entities in the military, government, civilian, and business sectors.

The cyberespionage campaign has been active since at least August 2025 and appears to align with Russian state interests, although researchers cannot confidently classify it as a nation-state operation.

Cybersecurity company WithSecure discovered the activity in January this year and determined that its focus is on Ukrainian or Ukraine-related organizations.

/* */