Anthropic co-founder Tom Brown: Each leading LLM is now a once-in-a-generation level genius mathematician. Over the next 12 months, LLMs will reach the same level in every key scientific field. At 16:02 https://t.co/xszDLboTTs
If everybody has it, nobody has it."
NYU Stern Prof. Aswath Damodaran points out that deploying AI won't automatically boost profits. If every company uses it the exact same way, there is no competitive advantage—only price wars and profit compression.
https://t.co/yComms7W2g
Be roughly right. If you're building a mental model of how the world works, you need to identify the drivers. If you demand precision on every variable, the "noise" of the data will drown out the "signal" of the logic.
In 2025, among major currencies, the Swedish Krona appreciated the most (16%) against the US dollar. That means the S&P, which returned 18% in USD, had a negative return, 1.18*(1-0.16)=0.9912, for a SEK based investor in 2025.
Keep in mind when vibe coding: every time you build something, you take on technical debt. Is it worth the technical debt you take on? Laurence Moroney. 48:45 https://t.co/lOqucaLj9H
Elon’s 5 steps for making things: 1) Make requirements less dumb. Question them. Person responsible for each. 2) Delete. Don’t assume it has to exist. 3) Simplify and optimize. 4) Accelerate cycle time. 5) Automate.
https://t.co/Nj89jV3zXQ
If nobody dares to take a risk, the organization stalls. Think of a penguin colony: if nobody dares to be the first to dive into the icy sea, the colony dies out.
Frederik Pferd, ex Chief Innovation Evangelist, Google
https://t.co/dulMOOibkK
Thinking out loud improves your decision making and problem solving. Strangely, if you ask an AI to think out loud it’s true for the AI too.
Jeremy Utley, AI & Creativity expert, Stanford
https://t.co/5YS6DNYamz
Trends:
1. Companies want good-enough agents to automate mundane day to day tasks. Not genius level.
2. IT departments, instead of buying SaaS products, see whether they can just build it themselves using vibe coding.
Ali Ghodsi, Databricks
https://t.co/ZWCdJd6Tkv
@alighodsi
In this recent interview, Vinod mentions a useful mental model for thinking about where we’re headed: in five years, AI will be able to do 80% of 80% of all jobs.
@vkhosla
https://t.co/jRvRp5kajI
Ray Dalio, interviewed in 2025. Retelling how in recent meetings with elected officials they say they agree tough actions are needed to address the deficit, but they can’t say that out of fear of not getting re-elected...
14:45
https://t.co/KsgcM7mMue
(2/2)
Former Singaporean leader Lee Kuan Yew, interviewed in 2009. Asked about the already then large US deficit, he replies ”in a sense this looks as if the elected leaders have lost their will to confront the people with the truth”.
47:30
https://t.co/YxGlKUbnqk
(1/2)
Can’t help but think that we are incredibly fortunate that Ray Dalio is persistently sharing his perspective on how the machine works and what’s going on in the world:
https://t.co/fEIj6JVxJi