The author shares their experience with a new agent app called bb, which they find superior to the previous t3 app due to its quick setup and extensibility. They also discuss the trend of extensible software, mentioning tools like Pi and its use by companies like Prime Intellect.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
As AI agent apps rapidly iterate, an industry observer's switch from t3 to bb reveals users' real need for extensibility and model flexibility, sparking thoughts on the rise of the 'builder' community.
In an August 6 article, the author revealed that they had just downloaded t3 as an 'all-in-one' agent app last week, but quickly switched to a new app called bb. This quick switch was not accidental but based on a comparison of real experiences: bb takes only about 4 seconds to set up on mobile, while t3 was 'a bit finicky.' This efficiency difference became the direct reason for the author's choice of bb.
bb's core appeal lies in its extensibility. The author emphasizes that bb not only supports mainstream agents like Claude, ChatGPT, Pi, Cursor, and Factory, but also allows users to build plugins by requesting agents, for example, the author asked Prime Agent to develop a Factory plugin for bb. This 'self-extension' capability is seen as a key feature of modern agent apps.
The author draws an analogy between bb and Pi, seeing both as embodying the idea of extensible software. Pi is described as 'one of the best coding tools,' and its minimalist design makes it more efficient than other tools in task execution. This minimalism has attracted companies including Prime Intellect to build their own agents on Pi, such as Prime Agent—a self-improving tool for coding and long-horizon tasks.
The author argues that traditional IDEs (like VSCode) are outdated, and the modern 'IDE' should be a desktop agent app that can be extended according to user needs. This view reflects a personal judgment on the evolution of development tools, but the author provides no specific data, relying only on personal experience and industry observation.
The author explicitly states that they cannot accept using an app all day that doesn't allow switching models and agent frameworks, because different tasks may require different models (e.g., GPT and Claude each have strengths). This need for model flexibility is a significant factor in attracting users to apps like bb.
This view is a personal preference, but it may represent the needs of a segment of advanced users. The author provides no statistics but emphasizes the value of integrating multiple models in a single app.
The author predicts that future software will take the form of 'widgets,' such as to-do lists, email, subscription readers, and larger tools (like the author's Twitter bookmarks and trends tool). Users should be able to add features on demand within their primary work tools to adapt to changing needs.
This prediction is based on the author's observation of the 'builder' community but lacks systematic evidence. The author believes that AI lowers the barrier to building tools, enabling non-developers to create personalized tools, thus driving the expansion of the 'builder' community.
The author quotes a view on X: 'You can't make the masses want automation processes when they want to hang out with friends and not work.' The author fully agrees, believing that agents have not yet been widely adopted by non-developers. However, he firmly believes that the 'builder' community will surge, defining it as 'anyone who wants to use agents to build tools for themselves or others.'
The author reflects on a decade-long belief that no-code tools were too clunky and limited, while AI greatly simplifies the building process. He likens learning to use AI to learning a language, where immersion accelerates mastery. This view is a personal inference but reflects optimism about AI's ease of use.
This article is a personal opinion and product experience sharing. All descriptions of product features and performance are based on the author's personal experience, without third-party verification or specific data. Industry developments (such as Bending Spoons acquiring Airtable, Jeff Dean's departure, etc.) are relayed by the author without primary sources.
AI agent apps are evolving from 'all-in-one tools' to 'extensible platforms,' with users' need for model flexibility and customization becoming increasingly prominent. Although the mass market has not yet exploded, the rise of the 'builder' community may be a key force driving the widespread adoption of AI applications.
Primary report
Primary source