X / Twitter
Claude Code Engineer Thariq
Thariq announced a major Claude Code UX upgrade: a renderer entirely rewritten using a virtual viewport. Users can now scroll output with a mouse while the prompt field stays at the bottom, and many longstanding small issues have been fixed. He also previewed `/buddy`, stressing that it is a real feature, not an April Fools' joke.
https://x.com/trq212/status/2039453692592873587
https://x.com/trq212/status/2039201498996035924
Anthropic Claude Code Engineer Cat Wu
Wu shared a practical workflow: develop ideas anytime with Claude Code in the Claude mobile app, then seamlessly continue in the local CLI, which she calls “teleporting sessions.” She included the Claude iOS download link to encourage mobile AI work habits.
https://x.com/_catwu/status/2039421527935033854
Replit CEO Amjad Masad
Masad demonstrated how Agent 4 turns Replit into something like an operating system: users can customize the platform indefinitely through skills, making it an extensible agent platform. He also reposted something he called a “new psyop,” presumably a humorous comment on an AI topic, and praised a demonstrated project as “Sick!”
https://x.com/amasad/status/2039429759344730549
Box CEO Aaron Levie
Levie said the ultimate limits on agent-driven productivity will be security, compliance, governance, people's capacity to review agents' work, and their ability to ensure alignment with regulations and organizational policies. Using April Fools' Day as a pretext, he also previewed an upcoming product or feature, saying “my team is going to kill me,” but that its official launch was close.
https://x.com/levie/status/2039216522028257549
https://x.com/levie/status/2039414479084278239
Every CEO Dan Shipper
Shipper highlighted Linear as evidence that SaaS is alive but must become agent-native. He considers it an example of a company founded before AI successfully transforming its product for the AI era. He also recommended a tool called proof and said he has been a heavy Notion user since meeting its team in 2018, considering it particularly well suited to writing.
https://x.com/danshipper/status/2039357127903350960
https://x.com/danshipper/status/2039149696166318320
Zara Zhang
Zhang officially released the “Follow builders” skill, a carefully curated tool covering 25 high-quality X accounts and podcasts. It helps users follow fast-moving, high-quality developments from AI practitioners, with a direct skill link provided. It is designed as a daily information feed with a high signal-to-noise ratio for people following AI builders.
https://x.com/zarazhangrui/status/2039368866741277074
https://x.com/zarazhangrui/status/2039368921959293126
Y Combinator CEO Garry Tan
Tan argued that the old rules are gone. What matters now is building things people actually want and using tools available to everyone; willingness and execution matter more than access to resources. He disputed how San Francisco's “overpaid executive tax” is described, calling it effectively an eightfold penalty on companies, calculated on a gross basis and accusing the media narrative of deliberate bias.
https://x.com/garrytan/status/2039131255531512266
OpenAI VP of Science Kevin Weil
Weil shared progress in AI mathematics: models are solving more open problems, while their proofs become more elegant as capabilities improve, marking a qualitative change in mathematical reasoning.
https://x.com/kevinweil/status/2039200605672284572
Linear Head of Product Nan Yu
Yu discussed “negative contribution” through a DotA analogy: a player who repeatedly dies both removes a teammate from play and gives the opposition resources. This “negative resource” effect exists in product teams too. He also emphasized that frequent interruptions from PMs or salespeople asking engineers about product details create systemic inefficiency. He demonstrated Claude Code reading code to answer such questions directly, avoiding the cost of manual communication.
https://x.com/thenanyu/status/2039490349941526770
https://x.com/thenanyu/status/2039468540982337550
Swyx
Swyx created oh-my-codex and wondered about an apparently missing open-source project: why hasn't anyone built an open-source Dropbox alternative on Cloudflare R2? He considers the need so obvious that someone must have done it and hopes to hear who. He also shared a famous line from JFK's moon speech for motivation and noted that Codex usage did not grow in March.
https://x.com/swyx/status/2039435085305585742
https://x.com/swyx/status/2039141201753071781
https://x.com/swyx/status/2039472186600427903
Podcasts
AI & I by Every — If SaaS Is Dead, Linear Didn't Get the Memo
Linear co-founder and CEO Karri Saarinen joined Every CEO Dan Shipper for an in-depth conversation. Their central point: SaaS is alive, but must evolve to become agent-native to survive the AI era, with Linear an exemplary case of this transformation.
Key takeaway: Linear resisted rushing out a chatbot early in the AI wave, spending two years truly understanding workflows before building an agent-native product.
Saarinen admitted that when GPT-3 arrived, Linear experimented internally with a chatbot but soon found it “not very useful.” His design background informs his approach of understanding before acting. Rather than copying others, he repeatedly asks where users actually need a capability and what the workflow is. This helped Linear avoid many unproductive early AI investments.
The real shift came when Linear positioned itself as an “orchestration hub for a multi-agent world.” Saarinen expects every person and company to have many agents. Linear's role is to provide intent, context, and task prioritization. When OpenAI released Codex Cloud Agent, it integrated directly with Linear because Linear already had an open platform and comprehensive API documentation. Customers including Coinbase and Brex even built internal coding agents and connected them to Linear as their agents' project-management system.
Saarinen takes a nuanced view of the “SaaS is dead” narrative around publicly traded SaaS companies. The direction is right: large companies' competitive advantages are weakening because inertia prevents rapid product reconstruction. But extreme claims that “people will vibe code their own CRM” oversimplify a more complex reality. Linear must also reconsider every product decision with a “Day 1 mindset,” without relying on past formulas for success.
Saarinen is skeptical of AI productivity measures such as token consumption or the number of agent-submitted PRs. Token sellers—AI companies—have incentives to promote these metrics, which do not represent actual value. He prefers traditional questions: has the product improved, and are there fewer bugs? Linear has a “zero-bug policy,” requiring every bug to be fixed within a week. Coding agents now attempt the first fix, engineers review it, and the entire process happens in Linear.
Personally, Saarinen still prefers drawing designs manually in Figma. Slowing down to consider every design decision has value, beyond speed alone. But product-team engineers and PMs have fully adopted agent coding tools, particularly for understanding requirements. He gave AI the company's product-philosophy documents to build an agent embodying “the Linear way,” rapidly analyzing large volumes of feature requests and helping prioritize them without manual review of each one.
Linear has approximately 120 employees, including 60 on the product team. In the AI era, this size enables agility and fast decisions, a core advantage that large legacy SaaS companies struggle to replicate.