What the people actually building AI said today. One page — a 2-min read.
Claude removed the five-hour limit for verified Ultra subscribers, while Replit introduced tools for maintaining hosted apps and serving government customers. Builders also focused on a harder problem: giving agents useful access to workplace systems without giving them unchecked control.
Claude removes the five-hour limit for verified Ultra subscribers
Claude renamed Max 5x to Max and Max 20x to Ultra, with prices unchanged at $100 and $200 per month. The five-hour limit is being removed for verified Ultra accounts, starting with most existing subscribers; newer accounts retain it temporarily until verification. Weekly limits remain unchanged.
Serval bets enterprise agents need tightly controlled access
Serval's Jake Stauch argues that better coding agents make approval flows, limited permissions and audit trails more valuable in large companies. “The boundaries are the product for us,” he says. Serval lets IT teams build automations with code-generating agents, then reviews and tracks what those automations can do.
Replit pitches autofix for hosted apps and introduces a government offering
Amjad Masad says apps hosted on Replit can become “Self-Maintaining Software” when autofix is enabled. He also introduced Replit for Government and shared a whitepaper on sandboxing agents to improve productivity while reducing risk.
GBrain Cloud aims to carry context across different agents
Garry Tan says GBrain Cloud lets users run the same context across agents with different strengths, including personal assistants and coding tools. His pitch is that people can switch between those tools without leaving their GBrain context behind.
Thariq challenges the default doom scenario for advanced AI
Thariq argues that people are poor at predicting how greater intelligence will affect the economy. He rejects the assumption that it must simply become expensive and displace everyone, and instead describes a well-aligned system that explains its work and makes individuals more effective.
Peter Steinberger argues for control over your own agent setup
Steinberger says companies can change what users are allowed to do in a managed agent setup, making a personally controlled setup appealing. He also says unpredictable inference quality had previously led him to recommend skipping intermediaries.
Source data comes from the open-source project follow-builders by zarazhangrui, released under the MIT license. Summaries are generated by an LLM from that project's public feeds, and the summarization prompts are adapted from it. Every item above links to its original source.
Summaries generated automatically. Read the original before relying on any claim.
