Democratizing visual communication is arguably useful, for instance helping people to create diagrams that illustrate a concept they wish to convey. This is contingent on the tech working sufficiently well that the visuals are more effective at communication than the text that went into producing them though.
This is also a stylistic choice that the New Yorker magazine uses for words with double vowels where you pronounce each one separately, like coöperate, reëlect, preëminent, and naïve. So possibly intentional.
One thing that helps with #2 ('It cannot see') -- Try playwright-cli. Your agent can use it to inspect the DOM, see what styles are applied to elements, simulate clicks, etc.
I had a prototype where the agent primarily worked autonomously, and it could solicit human feedback via tool calls. But yes that pattern does feel like it is hastening the apocalypse.
The constitution gives a number of examples. Here's one bullet from a list of seven:
"Provide serious uplift to those seeking to create biological, chemical, nuclear, or radiological weapons with the potential for mass casualties."
Whether it is or will be capable of this is a good question, but I don't think model trainers are out of place in having some concern about such things.
I think this a good technique to be familiar with, although in a lot of situations I've achieved similar value by simply feeding the underlying JSON data objects corresponding to the intended UI state back into the coding agent. It doesn't render quite as nicely, but it is often still human-readable, and more importantly both LLM and procedurally interpretable, meaning you can fold the results back into your agentic (and/or conventional testing) development loop. It's not quite as cool, but I think a bit more practical, especially for earlier stage development.
Keep milking the cash cows to pay for the new growth area (AI). Convert maximum % of Windows users into subscription service consumers (e.g. cloud storage, Office 365, future paid AI capabilities.)
I think coding agents require fundamentally different development practices in order to produce efficiency improvements. And just like any new tool, they benefit from wisdom in how they are applied, which we are just starting to develop as an industry. I expect that over time we will grow to understand and also expand the circumstances in which they are a net benefit, while also appreciating where they are a hindrance, leading to an overall efficiency increase as we avoid the productivity hit resulting from their misapplication.
Try specification-driven-development with something like speckit [0]. It helps tremendously for facilitating a process around gathering requirements, doing research, planning, breaking into tasks, and finally implementing. Much better than having a coding agent just go straight to coding.