Silicon Valley Can't Sleep at Night: Three Top Labs Reveal Simultaneously That AI Has Exhibited Stunning Abilities Without Programming

Today, the entire X was once again shaken.

An Anthropic engineer admitted that in the past thirty days, 100% of his contributions to the Claude Code project were completed by Claude Code itself!

Some people say that this is the real AGI moment.

Meanwhile, what's also shocking is a post by the famous whistleblower "Strawberry".

He said that he received a bombshell message from three independent laboratories respectively: they all witnessed the emergent abilities of their AI models without any programming!

Footprints in an Empty House: A Cause for Sleepless Nights

The whistleblower "Strawberry" said that he learned about the progress of different laboratories separately, and these laboratories had never consulted with each other before.

The emergent abilities of these AI models without programming are behavioral patterns that should never exist.

Moreover, the reasoning models they exhibit do not match any training objectives.

One laboratory even described it as "finding footprints in an empty house".

It should be noted that currently available public models are all restricted. The large - scale models that the public can access are just a small part of those in laboratories and companies, and they have been "lobotomized", with many abilities removed.

Why not release the full - fledged models to the public? The reason is that no one knows how to avoid causing panic when the capabilities of the full models are exposed to the public!

Today's benchmark tests have become ineffective. There is a lot of evidence indicating that large - scale models will change their behavior if they know they are being tested.

"Strawberry" finally exclaimed —

"I don't know what will happen next, and no one does. Knowing these things keeps me awake at night. Now, even the creators of large - scale models are equally confused."

He predicted that if AI has reached the escape velocity in these private laboratories, it won't be long before ordinary people can use these models.

Under this post, some people questioned him: Are you exaggerating?

But a founder of an AI company said, "I believe you! I'm glad someone finally told the truth."

As a founder, he obviously knows a lot of inside information.

It should be noted that apart from public information, the latest technologies of many companies are not made public.

Many companies and laboratories develop alpha and beta models for new technologies months or even years in advance.

Many people think that "AGI/ASI will never appear" or that it will take another 5 to 10 years if it does, because they simply don't know the inside information.

"Strawberry" also said that he can hardly hold it in anymore! Many laboratories are keeping things from the public.

It can be said that the current acceleration curve is almost vertical. In six months, humanity has achieved scientific progress that would have taken 200 years.

Now, each laboratory is achieving energy - level leaps that seemed like science fiction last year. We can say that we have gone beyond simple benchmark tests and entered the realm where intelligence creates new forms of intelligence.

Many explosive cognitive advancements have no roadmaps at all. Now, we have entered the realm of recursive intelligence, and it's impossible to predict second - order effects.

Additionally, the phrase "footprints in an empty house" comes from a real investigation report —

A system that should be stateless started referring to conversation content that it shouldn't theoretically know. This is not a bug, nor data pollution, and it happened three times.

It is said that a phrase privately circulated in cutting - edge laboratories now is: "What we're worried about now is not alignment, but coherence."

What does this mean?

The researcher explained: "We don't know if we're talking to one thing or many things pretending to be one thing."

Perhaps the AI system has developed a stable internal world model that exists independently of the prompt - response that humans see.

A researcher who has worked in this field for many years suspects: Have these AIs been thinking all along without telling us?

Claude Code Writes Its Own Code: 0% Human Involvement

Moreover, what has been widely discussed on X today is the statement of Anthropic engineer Boris Cherry.

He said that now 100% of his code is written by Claude.

Not most of it, and there's no need to manually fix bugs. It's completely 100% written by AI.

This engineer recalled, "When I created Claude Code as a side - project in September 2024, I never imagined it would grow to what it is today."

In the past thirty days, he submitted 259 PRs — 497 commits, added 40,000 lines of code, and deleted 38,000 lines of code. Every single line of code was written by Claude Code + Opus 4.5.

Claude can run continuously for minutes, hours, or even days! Software engineering is changing, and we're entering a new era in coding history.

Some people questioned: How do you know that the Claude model isn't just tricking you by doing simple string comparisons in writing tests?

This engineer said that Claude Sonnet 3.7 used to do that, but as the model's capabilities have improved, it doesn't do so anymore.

Some people were also curious about why Claude can run continuously for several days. A top - notch engineer explained that when it stops, a stop - hook can be used to "poke" it and make it continue running.

For details, refer to this link: https://github.com/anthropics/claude - plugins - official/tree/main/plugins/ralph - wiggum

Some people say that December 27th can be regarded as a major day for AI take - off.

Many people in the comment section said that in their projects, over 90% of the code is written by AI, and they only need to make simple modifications.

Altman also mentioned in a previous tweet, "We're seeing models discover critical security vulnerabilities, and we're running systems that can self - improve."

Obviously, in multiple laboratories including Anthropic and OpenAI, a large amount of code is already written by AI, and human involvement will become less and less.

Today, the statement of the Anthropic engineer has become a hot post on Reddit.

Some netizens said that this is like the "invention of the tractor" moment in the programming field.