X / Twitter
AI tutorial and interview creator Peter Yang
Peter Yang broke down competition among personal agents into product experience, distribution, work and personal use cases, and collaboration among multiple people, explicitly describing these assessments as his views as an observer. He is optimistic about Muse because the app is intuitive, Meta is promoting it heavily across its platforms, and the experience of a standalone app makes him question whether putting agents into existing messaging apps such as iMessage makes sense. As for ChatGPT, he believes its 1B+ users could still put it in the lead, with advantages in models, computer use, and voice capabilities, but serving both work and personal needs, and both businesses and consumers, within the same product creates experience design challenges. He has started using Grok Bot for work tasks himself, partly because the distinction between Work and Codex still confuses him. The direction he finds promising for Grok Bot is an agent version of Slack, where multiple bots work alongside people. Google’s advantage is that other personal agents rely heavily on Gmail, Google Calendar, and data from Google apps. However, the last time he checked, Spark was still only a secondary tab within Gemini, so he argued that Google should make it the primary experience and improve its usability. Apple has advantages in devices and its reputation for privacy, but in his view, releasing major updates annually and not actively gathering feedback on platforms such as X make Siri’s iteration appear too slow from the outside. Collaboration among multiple people is a key unresolved gap in his view: for example, it is not easy to add a spouse to a Muse chat to plan a trip, or smoothly bring a colleague into a ChatGPT or Grok Bot conversation. Most AI collaboration among multiple people still amounts to mentioning a bot in Slack. His current tool mix is Muse for personal matters, Grok Bot for cloud tasks, Claude for specific use cases, and ChatGPT for other needs. Faced with a constantly changing competitive landscape, he recommends making personal skills and files easy to move across agents and runtime frameworks, to avoid spending too much time switching tools.
https://x.com/petergyang/status/2101862331345154469
https://x.com/petergyang/status/2101865476145959373
Vercel CEO Guillermo Rauch
Guillermo Rauch shared an experience in which an agent investigated a rendering issue in a mobile in-app browser, and said agents stand out for their thoroughness in software testing and quality checks. After he pointed out a display anomaly, the agent kept working through reproduction, simulation, fixing, deployment, and verification, covering the full process from identifying the problem to confirming the result. One specific action was creating a temporary Vercel deployment and then testing it in an iPhone simulator to verify its behavior in that environment. He compared this sustained investigation with human developers, arguing that people are constrained by their energy and can fall back on assumptions or the hope that the problem has been solved partway through. This experience made him optimistic about the quality and performance of future software, based on the agent’s investment in verification and stamina in carrying out the work in this case.