1. $20 tier for Anthropic, OpenAI, and Google.

2. Use OpenRouter for some agentic work and for using GLM 5.3 for some security reviews.

3. Mostly the model harnesses (desktop apps for business and personal tasks and cli harnesses for coding). Use pi with openrouter.

4. My thesis is that one-time work (creative, coding, etc.) is different than on-going repetitive work--the work of the back office. I've built a platform for the latter specifically around centralizing agents while decentralizing orchestration. (https://sleeperservice.io/)

5. Waiting on a Duet invitation. But Dots is interesting to me. Use Gemini with Gmail to some extent.

Exactly! I called that the Operations Equation and wrote a tiny bit more here about it: https://zero2data.substack.com/p/the-operations-equation

Once you are inserting models into non-creative (and software development is creative work) it is imperative to start creating your own small evals. Not just to check how different models perform in that equation, but to catch model drift over time.

The moat isn't going to be the model (short of AGI). The moat is going to be what surrounds it. The harness and, I have suggested elsewhere, memory. The more personalized the AI is to you the more valuable it will be.

Also, as tech companies grow, a significant portion of the dev body count is for internal tooling. I was always flabbergasted at Twitter's headcount growth back in the day when it seemed like the only innovation increasing the number of characters in a tweet.

Right now we are in a weird spot where a lot of people are flailing around inside corporate America trying to figure out how to leverage AI. Few have enough vision to see what the companies of the future will look like and work towards that.

With how easy it becomes to write software, I'd say that both the model and the harness are already being commoditized. I don't see a reason to believe there would be moat in any of them.

Software engineering is a sliver of overall economic output. It is an input for tech companies who sell something else (ads, subscriptions, etc.). It is a backoffice expense for non-tech firms that do engineering.

I anticipate that the harness (and memory) will evolve far more outside of engineering. Muse/Dots are an example of that.

I've been writing about this thesis for the past year or so and here is a bit of a summary of where I think it is going: https://zero2data.substack.com/p/the-future-of-consumer-ai

Could you point me to something that their harness does exceptionally well as compared to their competitors, and why that can't be accomplished by a third party?

That was the idea that stuck with me the most from the novel. I did a bit of a surface level dig into what that would look like:;

https://www.startuplevelup.com/corporations-as-an-api/

What I think is missing is the lineage of an agent's identity. A public key maybe?

I see that as well. Sometimes I can see that they are prioritizing speed to print for breaking news (publishing to the web happens continuously vs the cadence of a daily newspaper) but often it is just being sloppy.

When most people are consuming _their_ news (as opposed to _the_ news) in the form of a thumbnail in an algorithmic social media feed, those copy errors matter less.

How is this different from Microsoft scraping to build Bing?

Honest question. There is a line in the sand somewhere apparently.

Under “conduct requirements” imposed by the CMA in June, UK websites are able to activate an opt-out to stop Google from scraping their content to power search features such as AI overviews - very similar conceptually to the news law passed in Australia.

https://www.bbc.co.uk/news/world-australia-56163550

Bing sent traffic to the original source. AI answers don't. That's the line. It's not complicated.

Linking to work, where ownership and attribution is clear and the owner has the ability to commercialise is a very different thing to “laundering” content through the model, quoting the midjourney developers here

> "We just need to launder it through a fine-tuned codex." [0]

[0] https://cybernews.com/news/midjourney-ai-images-art-lawsuit-...

Huge difference between building AI and a search index.

Why? In both cases the SaaS downloaded the whole web and derives 100% of revenue from content they didn't make.

Yes but the product is different. One is a shovel for you to dig, another one is an artificial intelligence that competes with or replaces the digger.

Bing isn’t re-selling you back the content it took.

They absolutely are; you pay by viewing the ads.

Every once in a while I would put Flight Simulator 2020 on my desktop and work on my laptop. I found that watching the digital world pass by while working put me in a slightly different headspace. Not sure it increased productivity but working while sitting in an actual plane does provide me a bit of focus.

Train ride videos on YouTube scratch some of the same itch.

there was a time when i would have a window open to a spacex web camera as they were building a launch site.

The Weber kettle grill is one of my favorite examples of quality in another way—function. It is built well but it also is such a wonderfully simple design that provides quite a bit of flexibility.

I have the Smokey Mountain which is amazing at holding heat but a total PIA to clean. Have one of their gas grills as well on a built in line, and it is fairly high quality also.

Nothing tops the kettle though.

I don’t want to read content written by AI.

A feature I would like on LinkedIn would be a 100% verified human content flag for users. LinkedIn can then do the scans and flip that to false for any users it catches posting AI assisted content. Let me filter that out of the feed.

If we don’t do something to maintain some standard of discourse, we lose intellectually and as a piece of our humanity.

LinkedIn is the only social media I use and it is on thin ice.

They don't want you to do that because you will realize how quiet and dead these spaces are becoming.

How do you think they will verify this?

With AI?

Under European law don't you have to mark AI generated content as such?

Is that self reported? Does it apply to content created by a U.S. based AI model? Posted on a U.S. website? Read by someone in the U.S.

I am not anticipating people self identifying slop.

When I moved from LA to St Louis many years ago, there wasn’t an IKEA yet. I was truly confused about where one was supposed to go to buy lamps. Any furniture stores that had them were selling ones that gave off strong grandma vibes.