visarga

> I can't wait to never see a reasoning button ever again.

I hope to see a memory-less mode that is not incognito. Want fresh contexts sometimes, but also want to keep the chats saved in history. Memory can spoil some creative work, it dials the model in too tightly.

AI helps you but also your competition, and gets factored in by investors while customers can use it to find better deals. The whole market is different even if a company did nothing.

Whatever you can cheaply do with AI is not a moat, if there is profit in there there will be quick imitation and competition will eat away those profits.

Models can be replaced easily, harnesses & AI tools too. And if cloud inference gets too expensive there are local models keeping the cloud prices hard capped.

Probably AI won't make anyone very rich.

Sent to solve one task, came back with half of it solved and 2 more problems.

> does anybody relate to this grown interest to plants and bird distinction with age

I have a similar interest but mine is about passively assimilating foreign languages while watching movies. Every time I grok a new expression it feels like a rush of dopamine.

What happens is that your competition uses AI, your investors factor AI in, and your customers use AI agents to find what they need and can switch providers with greater ease in the agentic era. So even if a company does nothing the economic environment around it changed. The AI boost is mandatory but produces few winners, it's mostly a scramble to swim harder just to stay in place. Benefits are competed away fast.

So it does not follow that companies can bank the savings from firing people. Anything AI can do for me it can do for my competition as well, and humans still make the difference. The big question in the AI age is "why pick me?" why hire me, why invest in my company, why buy my product, in a sea of similar products made by everyone. A differentiation crisis accentuated by AI.

> Why does this “just work?” Nobody really knows, but it clearly does.

We know language has to be learnable by every human, so it needs to be really independent of any specific brain development particularities. If it was not accessible to babies there would be no more language next generation.

The system prompt part is surely cached across all users.

Many people make the mistake to think LLMs are stochastic parrots, but they are more like pianos, you play the keys and they make music, if you don't bring no new spices to the soup no wonder it comes out bland.

Don't just focus on the model, half the turns in a session belong to the human, it's an instrument that depends on how well the user wields it. It does not need to produce the novel insights, just to support the user to have them.

Best interface in my opinion is a git tracked folder, files as state, agents coming in and doing work. It's not just a chat interface because in chat mode everything flows like water under the bridge. It is files as agents, where you get the agent to write a task in a file, and then it updates the file like a blackboard as it performs work. Same task file can be seen by worker agent, multiple judge agents (for plan and implementation), documentation agents and reflexion agents (improve workflow). I would also track every user message in a chat_log.md for reflexion agents (what is wrong with the workflow) and intent review (is the agent still following user intent?).

As long as it is not substantially similar to training set it should be OK to reuse ideas. Ideas should not be protected by copyright or we can't create anything. We should not accept "vibe copyrights".

Everyone treats the human user of the AI as furniture, but they steer the whole process into unique directions.

> The output of it is also a derivative work,

A short session is just retrieval, a long session is always unique. The more the user writes the more it diverges from any content in the dataset.

You can automate work but never owning the consequences. AI tasks emerge from human contexts, they perform work in the context and finally the outcomes collect in the context - gains, losses, risks, costs. So AI is great but it needs our skin for the start, middle and end of a task.

I really hope this is how we are going to adapt to LLM era.

> if a polished appearance (even if it is mega plastic) is no longer a sign of quality, then what is? What signals do I need to look for now?

AI produces apparent uniformity, it makes it harder to distinguish the quality of the underlying process. Same thing happens in all fields, we are all harder to differentiate, then why would you pick me over the others? this is the hard question we are facing now, and it's a desperate situation to be in. Not just artists and people looking for a job, but also companies to their investors and products to customers. All harder to tell apart.

I think this drives a lot of the AI craze. If I can't own my niche anymore, what is my new specialization? We have no answers yet. We have seen this kind of total competition in a domain with cheap imitation in SEO and slop writing.

> So, continuing to profit--forever--from someone's else work, at scale, without their prior consent, is fair use?

Yeah, you are right. Have you been paying your dues to the authors of your math books in first 4 grades? I think 15% of your wages as an engineer would suffice. These kids continue to profit for years, and they are so many. Gotta pot a stop to that IP theft.

In my mistake I thought copyright was about copying rights, not paying for using ideas themselves. If just being downstream from a copyrighted work is infringement even without substantial similarity, then it's more like patents that expire in lifetime + a million years.

What do you mean pirating? They don't even distribute the originals, LLMs are not for replication, we already have copying and internet for that.

Why would we use a multi-billion parameter model to copy text? If we wanted the originals it would be easier to find them free, pirate or pay, if we use LLMs it is because we want something ELSE.

And caring about content rights in a world with limitless content and scarce attention is a mistake, it was never the content that was scarce in the last 20 years.

What you are saying leads to "pulling the ladder behind you" effect on creativity. It's impossible to protect more than substantial similarity and still allow creativity to exist.

> There needs to be a royalty payment based on if the AI regurgitates existing ideas.

So, by that logic, you need to be paying every time you regurgitate any of my ideas. Or anyone else's. Copyright now protects abstractions and vibes. Substantial similarity test be damned. Nobody can write stories about wizard schools, the idea is taken.

Right now I tried "What is digestion?" -> "Fable 5's safeguards flagged this message. Our intentionally broad safeguards deliver more capabilities but can also flag safe coding, cybersecurity, and biology tasks. Send feedback or learn more."

I don't think you can 100% ensure your queries have absolutely no biology and cyber keywords inside. No matter how harmless, they always trigger. People complain Fable aborts even when they try to make a login page for showing "username" and "password".

This makes models like Fable 5 impossible to use in any serious agentic task, because you can't even guarantee the model, which is a basic thing you need to build on.

> The GCP team wants their slice, the other team wants some otjer slice, and so on. Everyone wants some crap for their promotion package.

I got a Google One plan for Gemini, but it came bundled with YT Premium lite, and that somehow made it impossible to renew YT Premium for 30 days. I suspect different teams stealing customers from each other.

More recently it's "Roasting your commodities".

That human might have used AI. You can never know. Hand fixed AI output, human just polished the corners? Light rewording of a full text written by hand, because the author is not confident in their writing? Actual human text, but after researching with AI?

One of my policies for agentic coding is to spend much effort in developing tests, coded tests not LLM based vibes. My projects have around 1:1 LOC between code and tests. Tests are like skin, when the skin is pricked it hurts, agents need to feel pain too.

OP's idea "everything is a text file" is good and I use it too. My plans are saved as task.md files, numbered and named. Work items are checkboxes inside the file, closed work items are checked and a comment is added on the same line to provide feedback about the implementation.

I also keep a current-state-of-the-world document, it should be <20KB of text, keep the essential decisions and intents. Loading it allows resuming in <30s.

Something I never saw anyone else do - I save all user messages in a chat_log.md file which is referenced for intent alignment and state recovery. I consider the chat log on the one hand, and coded tests on the other hand as the two walls, the agent works in the mid section between them.

https://horiacristescu.github.io/claude-playbook-plugin/docs...

> There's no possible experience of the world there. No input other than a bunch of imperfect labels we created for stuff.

The brain too sits locked inside a bone box and only gets a bundle of unlabeled nerves connecting it to the outside. How can the brain could possibly experience anything, it only sees patters and patterns of patterns never the real thing?

I get annoyed when I see AI telltale signs too, but for example in my case I type 10x more than the final piece, is it really AI generated or just reworded? I don't use AI to fix my comments, just when I want to format article size pieces I post on my blog.

How about if you did part of your research with AI, then typed the text yourself?

This fits perfectly with my philosophy which says cost constraints determine internal structure in a system, structure and cost evolve together. While philosophers like to ignore costs and come up with "ideal conceivers" as if they would inherit our limited concepts but just have unlimited computational budget.

The "Chinese Room" has the rule book baked in from the start, costs were hidden. When you remove cost from the system of course you can't see the semantics.

My own experience is that Opus 4.8 has an adversarial-teacher voice, unsolicited grading as if I submitted an essay for grading, declarations about the "real" issue, and constant "honest notes" self grading its own responses even before it answers. I can't stand its tone. We can't have a normal chat.

While Fable reverts to Opus for simple questions like "What is digestion?"

Can't stand NYT ever since they subpoenaed for millions of chat logs from OpenAI, trashing user privacy for their own goals on a massive scale.