The solution could be to cancel my AI subscription

I'm trying to list all the wonderful things I built with artificial intelligence:

Except... SaaS There is hardly any other useful, and I do not want to preserve any of them. I inadvertently run a news media, which is undoubtedly a burden. Of course, it helped me “learning artificial intelligence tools”, and I used many of them, but I didn't need them. I cannot sustain any of them, whether in terms of time, commitment, faith, attention or willingness to spend money.

I didn't mean to build these things. Usually.Claude.Start Session

Similar to writing a fast script for X

In an hour.

Result is not a fast script for X, and usually it's not mine.

Problem solved, no matter what itches in the first place.

On the last point, this technology is very impressive. It's a thermonuclear multiactivity amplifier, which I see in every adult friend of mine. People run three screens at the same time, doing completely irrelevant “projects”, and they have little hope of continuing, and so little commitment to results, that time is clearly wasted.

Recently, at least one screenshot is sent every month.

They're developing great tools. I'm like, wow, it's really something.

The author is clearly proud and passionate. I tried not to ask, but I...

Always wondering where you're going to sell it?

Because when it's a problem

When asked about the engineer, the answer came from LLM There was no change before it existed.

I've recently been interviewed, and when it comes to the use of artificial intelligence,

That's the same answer. We're easy on it. There's five at most.

They run the agent's room.

I felt a little nervous at once.

My stomach.

After months of using Claude, I had a blurry sense of effect. I then reduced my subscription to Pro because I believed that quota restrictions could reduce overuse. Then Claude went through a bad period of service and I moved to Cordex. Codex's CLI is much better than Claude's and is clearly faster. Use starts to pick up.

This technology, which has been refined, is truly impressive. It was required to conduct zero-shooting and comprehensive testing of the deep Austrian grammar in deep language. The tools that exist today must not be promoted as wisely as the focus it requires.

Almost every supplier and every tool intends to do exactly the opposite: more use, more tokens, more output. Ask a simple question. ChatGPT Yes/no, it is clear to you that it is hard to include relevant follow-up questions to promote excessive interaction.

Throw 10,000 untested LOCs in 5 minutes Python./JS chaos doesn't help anyone. The idea of this happening in every business environment at the same time is alarming.

One of my early artificial intelligence experiments, exploring artificial intelligence as a lens of Marshall McLuhan-style thinking, was to connect voice recognition to the conduit that generated blog posts at the other end, confident that it would encourage me to capture my ideas. All I need is to press the voice comment button on the Telegram channel, and then pop up the Opus-format post.

The output is uncontrolled garbage. As efforts are eliminated, commitments are eliminated, and with commitments the focus disappears, and with focus, any meaningful product disappears. High-quality writing is not simply a session that is projected by lens in English: English is low bitrate noise, and high-quality writing attempts to capture high bitrate information with better concepts, which should have been obvious before I started.

I thought about reusing the pipeline to capture private notes, but I don't need private notes. It subverts the natural process of noise being forgotten. This is just an increase in overuse of tools.

It follows that as long as quality matters, I believe that handwritten writing will never become obsolete.

I doubt that the answer is “better models” or “better tools”. Carl Newport links this to pseudo-productivity:

According to speakers, digital productivity tools, including artificial intelligence and e-mail, often result in a “digital productivity paradox”: They make individual tasks faster or easier, but they may make knowledge workers more busy, more distracted and less productive overall. Research cited by him showed that artificially intelligent users spend more time on e-mail, news, chat and business management tools, while spending less time on focused and uninterrupted work. His core assertion is that tools aimed at reducing friction usually increase the number of subsurface tasks and context switching, thereby weakening in-depth work and high-value outputs.

He explained that this was due to the fact that knowledge work was usually dependent on “false productivity”, in which visible busyness was seen as a representative of real value. Digital tools reinforce this by making people look more active: sending more messages, generating more drafts, attending more meetings and generating more work. In order to avoid falling into a trap, he suggested measuring real results, identifying real bottlenecks at work and separating in-depth work from shallow work so that digital tools supported meaningful progress rather than draining attention.

Other Organiser

These experiences have opened up a new understanding of the use of all tools, because behind all this is not a faster development = more applications or faster e-mail = more communication is an ideal goal. In general, it is about a unit of time in life and how it is spent meaningfully.

I do not currently know how to manage artificial intelligence, unless it is reduced, because a tool that produces cheap returns with minimal input and without friction can only be a burden, and achieving this may be the only real contribution of artificial intelligence to date.

David, May 31, 2026 Sunday 14:31:04.