Blog / WEEKLY #002
Weekly #002 | iPhone Duo launches; does Astra using a computer count as AGI?
Recording by itself produces no value; review does. When watch, training, and health data are automatically compiled into a weekly report, continuous logging and scheduled review finally connect into a closed loop.
Moving plugins from one agent to another, format conversion is no longer the hard part; the real threshold is that each platform exposes different capabilities, and upstream keeps changing fast.
As models start acting on intent rather than literal instructions, the center of gravity in collaboration shifts from prescribing every step to stating goals and acceptance criteria. Which constraints still need to be written down explicitly is worth discussing separately.
How open a piece of software is to agents depends on whether it exposes capability interfaces, whether results can be inspected, and whether the host can carry the execution. From 3D modeling to editing tools, this line of judgment is taking shape.
The ceiling of what a life assistant can do often is not the model — it is where the information it can see lives: the same school notice is readily at hand in email, but hard to reach inside a group chat.
Personal agents have lowered the bar to "being able to chat", but what decides retention is not the chatting — it is the first useful task: scenario entry points, an execution process you can see, and a conversation the user never has to manage.
Whether logins, software, and files survive after a task ends is splitting agent execution environments into two generations. Persistent environments make scheduled tasks and long-running collaboration possible — and push cost and multi-machine management onto the user.
Models keep getting faster and code keeps getting cheaper to write, but the speed at which software reaches users has not automatically caught up. Between individual output and organizational delivery sits a whole set of steps that have yet to be redesigned.
A promo video can be composed from a musical score, vector animation, and a handful of generated images. Code handles the orchestration, image models supply the key assets, and whether it looks good still has to be judged by eye.
Accuracy is no longer the main bottleneck for voice input. What decides whether people stick with it is a set of unglamorous frictions: vocabularies, real-time display, how aggressively the text gets cleaned up, and system permissions.
Beyond leaderboard scores, a model's real performance depends on the runtime framework, the type of task, and how fresh the test data is. Using concrete tasks as examples, this article explains how to read evaluation results — and how to use them.
The same idea—"have AI make a game"—can cost wildly different amounts and produce wildly different results depending on how the task is expressed. One attempt at building a planet shows how engineering expression determines the size of the task.
Recording, editing, platform uploads, and the website transcript used to be separate manual chores. When the steps beyond writing code can also be handed to an agent, the most human-intensive part of the pipeline shifts from operating to deciding.
The basic execution capabilities of an agent are being taken over by managed services and mature frameworks. The question an application team has to answer is not whether to build in-house, but which responsibilities to hand over and which to keep.
After the prompts are tuned and the pipeline works, model behavior can quietly shift one day. This article breaks the "models got dumber" debate into separate questions: versions, routing, pricing, and compatibility.
Once model prices fall, the first thing to change is not the capability rankings but which tasks become worth handing to a model. This article maps how cheap models expand usage, and how they divide work with frontier models.
