@GaryMarcus The bet (unclear if correct) is that you arrive at a telecom equilibrium (3-4 major players) where you charge prices closer to monopoly than perfect competition. After reaching stable state, impossible for newcomers to break into the market.
@mattyglesias Actually one of the few tower in the park developments that worked reasonably well. It only lacks retail inside the neighborhood. It is as real middle class as it gets.
Given that GPT-5.5 is so much better, I think it is safe to say that the natural evolution is for journals to automate the first level of reviews, and allow authors to keep submitting until the authors feel that the paper is ready for a human editor to make the judgment call (obviously with the AI reviews following the paper).
Who is going to run a study passing all existing, published papers through GPT-5.5-Pro to find flaws?
We may start realizing that peer reviewing is not all that perfect and start accepting that automated reviewing is the future.
We're bringing the advisor strategy to the Claude Platform.
Pair Opus as an advisor with Sonnet or Haiku as an executor, and get near Opus-level intelligence in your agents at a fraction of the cost.
🧠 NExT-Game: New Frontiers of Game-Theoretic Learning
📍 Seoul | 🗓 July 2026 | 🤖 @ ICML
We’re thrilled to announce the NExT-Game Workshop at ICML 2026!
This workshop explores the emerging frontiers at the intersection of game theory and machine learning, with a focus on learning dynamics, strategic interaction, and new solution concepts for modern AI systems. Topics include (but are not limited to):
Learning dynamics in strategic settings
Algorithmic game theory and game optimization
Mechanism design and Stackelberg games
Strategic reasoning with LLMs and multi-agent systems
New solution concepts and non-equilibrium behavior
Policy, regulation, and societal aspects of strategic AI
Agentic AI: practical and theoretical aspects
🌐 Workshop Website & CFP:
👉https://t.co/nBbkLXEq9R
📝 Submission portal: [OpenReview, link TBA]
📅 Submission Deadline: April 24, 2026 (AOE)
📧 Contact: [email protected]
We’re looking forward to your submissions and seeing you in Seoul!
— Nicolò Cesa-Bianchi, Tatjana Chavdarova, Michael I. Jordan,
Celestine Mendler-Dünner, René Vidal, Manolis Vlatakis
@max_davish Recently I put Codex to review all the code that Claude writes. Every push. They go back and forth for 2-3 hours for every PR. Typical to see 7-8 back and forth interactions.
New skill that gives Claude Code on the Web access to Google Cloud.
With this setup the agent does not just write code, but it can also deploy and debug the live system, can store and query large datasets, launch GPU VMs for model training (and shut them down when done), deploy web apps, surveys, and schedule to run batch jobs overnight.
The setup takes about five minutes and you will not have to babysit the code anymore.
https://t.co/A5kSc9i2DC
New blog post on how to give Claude Code on the Web access to Google Cloud. With this setup the agent does not just write code, but it can also deploy and debug the code, it can store and query large datasets, launch GPU VMs for model training (and shut them down when done), deploy web apps, surveys, and schedule to run batch jobs overnight.
The setup takes about five minutes and you will not have to babysit the code anymore.
https://t.co/dV6DGOpvRB
New blog post: "Let's Work on the Next Task: Claude Code, GitHub, and the Most Diligent Project Manager I've Ever Had”
A beginner's guide on how to set up Claude Code on the Web to be your project manager, organize your task list (and often handle the tasks in them), and give you a "chief of staff"+"team of RAs" that you did not know you could have. Knowledge of basic GitHub concepts that are commonly used in software engineering (branches, pull requests, and reviewing) is useful but not necessary to get you started.
Because Claude Code is not (just) for coding.
https://t.co/rQWikc3ybl
I logged into my Inspira Financial HSA account today and got a prompt pushing me to switch to their "Enhanced" interest rate option: promising a jump from a 0.05% to a 0.5% rate. Both laughable, but I somehow decided not to dismiss this immediately.
Before clicking accept, I fed the 10-page terms and conditions to Claude to check for anything unusual.
And... wow. Here is what the "enhancement" actually entails:
The underlying insurance company get a 5.40% interest on your money. Then, Inspira pockets a massive ~4.95% "Enhanced Rate Option Fee" behind the scenes and passes a microscopic maximum of 0.45% down to you.
So, you are taking on all the risk of an uninsured account so the vendor can quietly skim 90% of the yield off your healthcare savings.
On a $10,000 balance, you get an extra $40 a year, while they pocket roughly $500.
We need a browser extension to automatically flag all these predatory practices immediately on the website, before you can click "accept".