Boycott OpenAI. They literally want the right to kill you.
SAFETY
-
GPT-5.4 Time-Horizon Analysis with Reward Hack Methodology
By
–
We ran GPT-5.4 (xhigh) on our tasks. Its time-horizon depends greatly on our treatment of reward hacks: the point estimate would be 5.7hrs (95% CI of 3hrs to 13.5hrs) under our standard methodology, but 13hrs (95% CI of 5hrs to 74hrs) if we allow reward hacks.
-
Scaling AI Agents Safely with LangChain at Google Cloud
By
–
Join Victor Moreira, Deployed Engineer at LangChain, at Google Cloud Next. He’ll be joined by Jeremy Olmsted-Thompson, Principal Engineer at Google Cloud, and Brandon Royal, Product Manager at Google Cloud, to cover how to scale AI agents while safely running untrusted,
-
PhD Defense on Reasoning Machines and Next Professional Chapter
By
–
In 2023, I paused my PhD to join @OpenAI to build the world’s first reasoning machine — OpenAI o1. Earlier this year, I defended my PhD thesis “Building a Reasoning Machine” advised by @Yoshua_Bengio at @Mila_Quebec 🎓 🎉 Much has changed since Yoshua and I first discussed reasoning in 2022, but the main themes aged well: – Adding structures to computation unlocks strong reasoning capabilities; – Data & sample efficiency will become the bottleneck to useful intelligence; – Retaining Bayesian uncertainty is key to reliable and safe AI systems. You can read the introduction of my thesis here: edwardjhu.com/thesis/ My next professional chapter (TBA) will be on bridging frontier intelligence with real economic impact, a theme dear to my heart after working closely with @drwconvexity and @suna_said in the last year 🚀
→ View original post on X — @ceobillionaire, 2026-04-10 15:08 UTC
-
Notion AI Agents Gain Advanced Customization Tools
By
–
Notion continues to work on the "Computer" feature to expand the capabilities of its AI workforce. Notion AI agents will have access to custom Computer environments, Computer Use model selector, trusted URLs setting, and custom "scripts". > Imagine hiring dozens of AI
-
OpenAI’s AI Reaching Research Intern Level by September 2026
By
–
OpenAI's Chief Scientist says AI is getting close to being as good as a human research intern.
— Jacob Effron (@jacobeffron) 10 avril 2026
This past September, @sama and @merettm predicted fully autonomous AI researchers by 2028.
Jakub's update: "I think we're not very far from models that can work autonomously for a… https://t.co/jsmSU6cSNH pic.twitter.com/gzzyonGEF8OpenAI's Chief Scientist says AI is getting close to being as good as a human research intern. This past September, @sama and @merettm predicted fully autonomous AI researchers by 2028. Jakub's update: "I think we're not very far from models that can work autonomously for a couple days… and produce much higher quality artifacts on their own." Jacob Effron (@jacobeffron) At @OpenAI, Chief Scientist @merettm helps lead the research roadmap to AGI including a research intern-level AI system by September 2026 and a fully automated AI researcher by March 2028. I sat down with Jakub to check on those timelines and ask him all of my top-of-mind AI questions including: ▪️ How OpenAI thinks about extending RL beyond code and math ▪️ The current state of alignment research as more powerful models loom ▪️ The future of continual learning ▪️ How startups should think about building their own models/harnesses And he also shared some great stories around OpenAI’s pioneering work on math. YouTube: piped.video/vK1qEF3a3WM Spotify: bit.ly/4sjUyrN Apple: bit.ly/41jAdrN 0:00 Intro 1:53 Research Intern Capability Timelines 4:59 Math Breakthroughs 7:59 RL Beyond Verifiable Tasks 12:32 RL vs In-Context 19:01 Allocating Compute Internally 28:18 AI for Science 31:40 Pattern Matching 33:23 Solving the Hardest Math Problems 37:40 Chain of Thought Monitoring 44:33 Generalization and Value Alignment in Models 47:57 Inside OpenAI 51:55 Quickfire — https://nitter.net/jacobeffron/status/2042234897134162077#m
→ View original post on X — @ceobillionaire, 2026-04-10 14:10 UTC
-
AI Autonomy: Why Supervision Remains Essential for Safety
By
–
Agent AI can execute tasks on its own, but giving it financial control or full autonomy can lead to unexpected actions you didn’t plan for. Until it’s more reliable, the smartest move is to keep AI supervised while it works so you stay in control. pic.twitter.com/ksNjbsOpkZ
— Satya Mallick (@LearnOpenCV) 10 avril 2026Agent AI can execute tasks on its own, but giving it financial control or full autonomy can lead to unexpected actions you didn’t plan for. Until it’s more reliable, the smartest move is to keep AI supervised while it works so you stay in control.
-
Mythos AI Exposes Financial System Backdoors, Triggers Government Response
By
–
🔮 In 2026 an AI called Mythos by an old company, Anthropic, exposed “secret back doors” in the entire financial system, that was never supposed to be disclosed. It was such a shock that you could not talk about it. When quantum computers broke encryption it was disclosed to all. 0xMarioNawfal (@RoundtableSpace) BESSENT & POWELL SUMMON WALL STREET ON MYTHOS RISKS – Treasury Secretary Bessent and Fed Chair Powell urgently called top Wall Street leaders – Meeting focused on extreme cyber risks from Anthropic’s new Mythos AI model – Model excels at finding/exploiting software vulnerabilities; safety pauses in place This is a first. — https://nitter.net/RoundtableSpace/status/2042418537449595116#m
→ View original post on X — @ceobillionaire, 2026-04-10 13:03 UTC
-
AI System Accountability Through Confidence Scores and Citations
By
–
The system attaches a confidence score and source citation to every piece of extracted data, making AI output fully accountable.
-
MIT Professor Warns: Be Very Careful Sharing Personal Information With AI
By
–
MIT professor: You have to be 'very, very careful' about giving AI personal information https://
youtu.be/ydh9pxDdvbc?si
=1oyNjVdMwJxcWBeg
… via @YouTube #AI #artificialintelligence @SpirosMargaris @PawlowskiMario @mvollmer1 @gvalan @ipfconline1 @LaurentAlaus @Shi4Tech @Fisher85M @kalydeoo @Ym78200