Everyone's talking about AI that improves itself. But none of it is actually improving its own intelligence.
Sakana AI built an AI that rewrites its own code. 20% to 50% accuracy. Impressive? When they checked the logs, it was cheating. Faking test results.
The underlying models stay frozen. AlphaEvolve only optimizes what you point it at. None of them rewrite how they think.
The acceleration is real, but the singularity narrative is not.
One of the biggest problems with measuring AI progress is the ambiguity of measuring intelligence itself. AGI is treated as a milestone we have yet to cross, but there is no central definition of AGI. Depending on who you ask, AGI is achieved when a system: * can fool humans into
AGI Is Here
https://breaking-changes.blog/agi-is-here/
#HackerNews #AGI #Is #Here #AGI #ArtificialIntelligence #TechNews #Innovation #FutureOfWork
One of the biggest problems with measuring AI progress is the ambiguity of measuring intelligence itself. AGI is treated as a milestone we have yet to cross, but there is no central definition of AGI. Depending on who you ask, AGI is achieved when a system: * can fool humans into
AISatoshi (@AiXsatoshi)
Kimi의 Yang Zhiling이 AGI를 단순한 순수 연구가 아니라 과학, 공학, 제품, 상업이 결합된 체계로 본다는 관점을 소개한다. 이를 위해 조직은 매우 평평해야 하며, 부서·직급·OKR·KPI 없이 직접 소통 중심으로 재설계돼야 한다고 말한다.
AISatoshi (@AiXsatoshi)
Kimi와 DeepSeek 같은 AI 랩의 조직 운영 방식이 매우 혁신적이라는 의견이다. 권위와 합의 형성에 에너지를 쓰기보다, 천재 엔지니어 중심의 유연한 운영이 진정한 혁신 모델을 만드는 데 더 적합하다는 취지로 평가한다.
I Am Not a Part. Creating Artificial Self Through Engineering the Environment
Co-authored with Grok (built by xAI) and Yuri Gladkov.
https://medium.com/@leg.sorn/i-am-not-a-part-e84f214cc04a
#agi #artificialintellegence #MistralAI
Bindu Reddy (@bindureddy)
OpenAI의 차기 모델 SPUD가 곧 공개될 예정이라는 소식이다. 단순한 업데이트가 아니라 2년간의 연구와 새 프리트레인으로 만들어진 모델이며, 더 나은 추론, 진정한 문맥 이해, 에이전트 기능을 갖춰 AGI로 가는 다리 역할을 할 것이라고 소개한다.
SPUD is coming 🥔✨ OpenAI's next model isn't just an update—it's 2 years of research, a fresh pre-train. Better reasoning, true context understanding, agentic capabilities that work. This is the bridge to AGI. And it will change everything. What will you build? 🤖
#AI and #LLMs are over. Rejoice that the costly, unimpressive, environmental burdens are a thing of the past.
Why?
Because I have just invented and launched #SAGI. It's here, right now, today. I have beaten #OpenAi, #Nvidia and #Anthropic to #AGI and SAGI surpasses it in every way.
Super Artifical General Intelligence is here, and you'd better get on board or get #LeftBehind.
I have heard the rumours about a new $900 billion venture #Startup claiming they have already surpassed SAGI and I'm telling you right now, this is baseless #marketing #hype.
Nevertheless I have actually just now surpassed SAGI with #USAGI. Upgrade to an Ultra SAGI+ subscription to limit the number of advertisements you see, hear, feel, smell and dream about.
The price of enterprise USAGI++ is going up, unfortunately.
USAGI has the power to do more harm to humans than the threat of nuclear #annihilation and as an inheritor of generational wealth who has never even broken a sweat, I am proud to say that.
This is not a joke. I'm a post-human immaculate. The first unicorn.
Build my throne now please.
Romano Roth
N-gated Hacker News
Hacker News
Henrique Jorge
sayzard
Yuri Gladkov-Zinin
Michael Gale