ako

Exactly, and a dsl gives you many options and tools to make the llm more reliable: a grammar enables deterministic syntax checker, error messages injecting correct syntax into the context, linting, terse examples in the skills for token optimization, and even the option to add the grammar to the llm to help it decide what next tokens are acceptable.

For open source llms you can give your grammar to the llm to help it decide what valid next tokens it can generate.

They don't need to, they can use tools to validate their assumptions.

I think the hard part is not the syntax itself but the shift in thinking: instead of procedural state manipulation expressing the desired end result in declarative set based relation algebra. I see developers struggling with breaking down complex queries in (inline) views / CTEs, thinking they need parameters, when things can be expressed as a queries on another query. Complaining about the lack of reusability, but not knowing about views.

Time to start using plsql, ADA with first class support for embedded SQL.

What is an original image? A good photographer can also create a completely different impression. And it's hard to get rules applied, suing is often too expensive, for consumers, and governments need to reduce costs, so don't have the funding. It will not be enforceable enough that people wont try tampering with their picture.

If you make it a technical solution, e.g., browser plugin, it become an economic opportunity that can create money instead of cost money.

Instead of fighting the use of AI for home interior picture, it might be more useful to have an AI that can correct the fabricated images. If the listing includes room sizes, an AI should be able to give you more realistic images. Maybe a browser plugin that makes all content honest?

You could use an external keyboard?

It’s expensive to build (many rules, not enough construction workers, expensive materials), expensive land, nitrogen policies, economic crisis 2008-2013 when there were no buyers for new houses caused a backlog of new houses.

I assume management thinks this will lead to better documentation practices and standardized processes so it becomes easier and cheaper to introduce new employees. In practice the opposite happens, employees get scared for their jobs, hire bad new employees so it's the new people that will get PIP-ed.

Sure, they’ll say whatever is needed to sell the products they have.

I think the Appstore was planned all along, just did not fit in the first release, so they adapted the launching narrative to: "the browser is enough for all 3rd party software".

Apparently not enough people want what you want, that is also democracy, accepting that things other people want can be prioritized,

Yes, you get to say what you want, but that doesn’t mean you get what you want. With millions of people all saying something different, nobody gets exactly what they want.

Most relevant question in today's world seems to me how to create incorruptible governments?

True, maybe bundle it with an OF account?

As far as i know Claude Code also uses LSP and tree-sitter to find things in your source code.

People "happily" spend that on drugs, tobacco, and alcohol, so why not on porn.

Makes me think of Peter Tosh: "Anything you can do, I can do it better, I'm the toughest".

Tool use score is 17.4% that seems really low, what does that mean?

I think Amazon is doing ok as the cloud where most customers run their LLM. I think a lot of companies are using e.g., Anthropic models on Bedrock so it lives inside their AWS cloud.

It’s better, it’s useful even for those who don’t have a deep knowledge of computers. I’d expect more AI users than programmers, than ms-word users, than excel users.

Probably over a year ago, when I first saw reasoning in action in a debugging session: it generated some code, ran it, could not explain the results, then said “let me add some print statements to debug”, reran the application, read the logs, and then stated “now I understand why it’s not working”. Plan, do, check, act in action, AI engineering its own context, and generating the missing information.

Some suggestions for naming this, depending on how you organize ownership, redistribution, who gets to decide what you need: paternalism, welfare capitalism, neo-feudalism, technofeudalism, welfare state, social democracy, UBI-and-oligarchy, oligarchy, plutocracy.

We are already in a position where we could house and feed everyone, but we choose not to do so. Instead a few own almost everything, money gets spend on killing people, instead of feeding them.

If we choose not to do it now, i don't see how we can make UBI work in the future.

I don't see how you can relate this to communism. Sounds more like oligarchy/roman empire: a few own almost everything, but most have almost nothing and are being controlled with bread and games. Marxism is the opposite where everything is owned by everyone (in theory).

OpenAI may be better positioned than you think with Codex being able to drive your applications without an API according to Nate: https://www.youtube.com/watch?v=2d9ZmA-4QzU&t=1093s

I don’t care who wrote it, as long as it is a good message, communicating what the sender intents to communicate, well articulated. I prefer it above a self written piece that is hard to understand, contains unclear wording, and could be interpreted in an incorrect way.

Not everyone is gifted with the ability to write well, so using tools to achieve that is no shame in my opinion. And no, practicing more is not always going to get you to a higher, sufficient level. We all have our limitations that cannot be crossed by just practicing more. Practice does not make Einstein.

I quite often send messages where I later think “i should have phrased it differently, maybe it was misunderstood”. And often I’ll respond way too late because lack of time.

I feel AI is a tool that helps me communicate better, and I expect that holds for many others as well.

Not understanding that some feel more effective using tools is also a sign of lack of empathy.

Because it’s often hard to find the right voice? because it’s often hard to describe something in the correct way? Communication is hard, finding the correct tone of voice often takes me a long time, so using AI makes me more efficient in a busy day?

I’m using ai to answer questions, but I instruct it in draft what answers to include, what info to include from my llm wiki (second brain). Saves time to write a correct response, can easily refer to past conversations, but definitely not 100% outsources to AI.

Agreed, I think the builders are the wrong people to ask for this self reflection. Anything can be used for good and bad. A knife can be used to prepare food, or to stab someone. A drone can be used to attack a country, or to defend a country. A wooden beam can be used to build a house, or to build a cross to crucify someone. Same way AI can be used for good and bad. And it’s up to the person using the tool to act moral. Unfortunately this world is full of people that consider power and money more important than morals. For me, religion falls in this bucket, lots of “do as I say, not do as I do”. And on top of this, there’s very little agreement on what is good, and what is bad. Morals are quite fuzzy, and flexible.