X / Twitter
Aaron Levie, Box CEO
Aaron Levie again emphasized that software engineering jobs will not disappear. Agents are the greatest source of leverage technical people have ever had, making this the time when they can have the greatest impact, individually or at team or company level. He expects smaller teams with agents to perform most important work, while technical skills matter more than ever.
https://x.com/levie/status/2049333853777764495
Peter Steinberger, OpenClaw Founder
Peter Steinberger shared his CI/CD and agent automation workflow: whenever a commit reaches main, Codex automatically reviews it. If it finds problems, another Codex instance generates a fix PR, which the reviewer checks again, repeating for up to 5 rounds if new issues appear. He also integrated Codex review into Clawsweeper, supporting automerge and iteration until no new issues remain. A post about a product leak received over 3500 likes, demonstrating strong community interest in OpenClaw updates.
https://x.com/steipete/status/2049290265026773172
https://x.com/steipete/status/2049290741013262522
https://x.com/steipete/status/2049356949523730699
https://x.com/steipete/status/2049518771023360010
Josh Woodward, Google Labs VP
Josh Woodward announced file generation and export in Gemini: users specify content and format, and Gemini generates and exports it directly. Supported formats include Google Docs, Word (.docx), PDF, Google Sheets, Excel (.xlsx), CSV, Google Slides, Markdown, LaTeX, TXT, and RTF. It is live across all platforms, and the post received over 2400 likes.
https://x.com/joshwoodward/status/2049524400131838436
Kevin Weil, OpenAI VP of Science
Kevin Weil reposted an insight on AI and radiologists: their job is not reading X-rays but healing patients. If AI accelerates image interpretation, radiologists can treat more people. He considers this a framework for AI’s effects across professions: AI is a tool, while professional judgment remains central.
https://x.com/kevinweil/status/2049584046213378229
Aditya Agarwal, South Park Commons Partner
After extensive use of OpenClaw, Hermes, and similar tools, Aditya Agarwal shared a firsthand observation: most people claiming to love OpenClaw have not really used it. Agents remain developer products, far from smooth use by ordinary consumers. This measured calibration of AI hype comes from the practical perspective of a former Dropbox CTO.
https://x.com/adityaag/status/2049530878083379427
Amjad Masad, Replit CEO
Amjad Masad announced free Replit accounts for teachers and a 50% student discount to broaden AI coding education. He also humorously described users being “woken at midnight by a site outage” after Replit deployments launched, suggesting continued expansion of hosting capabilities.
https://x.com/amasad/status/2049595212889247994
Claude, Anthropic’s Official Account
Claude shared hackathon winners. Chilean developer Benjamin Torralbo, a carpenter’s son, built MaestrIA to identify household damage from photos, provide a diagnosis, list local parts prices, and draft messages to nearby tradespeople, winning the “Keep Thinking” award. French team ARIA reads equipment manuals and automatically creates fault tickets incorporating past repair experience, winning “Best Use of Claude Managed Agents.”
https://x.com/claudeai/status/2049523908584386582
https://x.com/claudeai/status/2049523910173966597
Podcasts
AI & I by Every — How Stripe Is Building for an Agent-native World
In this Every episode, Dan Shipper speaks with Stripe’s head of data and AI, Emily Glassberg Sands, about the agent economy’s implications for Stripe and the broader internet economy.
Emily identifies a fundamental assumption breaking down: the internet was built around humans as its primary participants, but agents are new actors. She describes three levels: humans remain in charge but interact through AI interfaces; agents act for humans; and software interacts directly with software. Together, these shifts demand redesigning foundational infrastructure for payments, billing, fraud detection, and identity verification.
Stripe’s data provides quantitative signals: LLM traffic to its documentation has increased 10-fold year over year, making machines important users of developer infrastructure. AI-native companies’ median time to $30 million ARR has fallen to around 18 months, placing them among history’s fastest-growing companies.
On business models, Emily predicts a sharp decline in B2B seat-based pricing within six months: charging by seats becomes absurd when agents make each developer 10 times more productive. Consumers, however, will retain fixed monthly fees because individuals generally dislike variable charges.
Stripe Link now has approximately 250 million consumers, and 58% of Lovable payment volume runs through it. Emily described evolving authorization for agents: rather than giving them credit-card details directly, use “delegated authorization with guardrails.” Consumers decide which agents may request credentials, under which conditions and limits, and whether pre-purchase approval is required, all managed through Link.
Stripe has also introduced machine-to-machine APIs optimized for agents and structured Markdown documentation specifically for LLMs, helping agents understand and use its developer tools.
https://www.youtube.com/playlist?list=PLuMcoKK9mKgHtW_o9h5sGO2vXrffKHwJL