luckydata

speaking to a child at their level is the best way to keep them from speaking well. I never did it with my son and it didn't hold him back one bit. Everyone remarks his incredible vocabulary and language skills for his age. IMHO holding back with kids is an anti-pattern.

In an immersion context I can get to conversational fluency in about 3 months and to complete mastery in a year. I've done it twice, once for Spanish and once for English. A few things in my approach helped me move quite a lot faster than my peers:

1. I would carry a mini-dictionary with me EVERYWHERE. Anytime there's a new word, I would ask a local to teach me how to pronounce it and then make sentences with it while I was walking around. CONSTANTLY.

2. accent and good basics help more than a vast vocabulary: when I went to spain for the first time I would hear in the metro the famous male and female voices saying "proxima parada... something something" and I would repeat that sentence trying to imitate the pronunciation and rythm to get used to "sounding spanish". That helped a lot.

3. date a local: in spain I was dating this girl that was a journalist and from a pretty conservative family. She was very afraid that I would put off her family by being a foreigner and not being able to pronounce things correctly or making grammatical mistakes so she would correct me on the spot EVERY TIME I said something wrong. I dind't mind it and it worked like a charm. Years later I met my American wife that wasn't nearly as concerned about my pronunciation in English so my accent is not nearly as good as in Spanish, but I definitely learned the language, went from being barely understandable to business meeting in about 4 months.

3. Watch tons of movies with the original subtitles (for example spanish movie with spanish subtitles) to understand how people pronounce certain words. DO NOT limit yourself to learner materials, you won't learn a thing. Find something you enjoy and just dive in, you'll learn a lot quicker that way.

Dedication and systematic work is all you need to move pretty quickly, the human brain is wired for language, if you feed it what it needs it will do the work for you.

new models are always magical, let's see how it feels after the cost cutting measures get implemented in 2-3 months.

your pricing is nuts. let me know if revenue are good, I'll copy you.

does this approach actually work? is it more reliable than plain mcp calls?

I downloaded the app and it failed at the first screen when I set up the models. I agree with the spirit of the blog post but the execution seems lacking.

yeah but it's hard for others to know what you're up to, you force everyone else to do investigations and waste time. Just open a ticket and be a good team mate.

have you ever witnessed how sometimes Gemini makes multiple attempts at writing a file only to give up and start chanting "I'm worthless...".

That's tool use failure :)

Gemini is great, it's just incredibly clumsy at tool use and that's why it fails so often in practice. I'm looking forward to the next version, it will for sure address it, it's a big issue internally too (I'm a recent xoogler).

I put together a small harness to keep AI-assisted coding under control. It’s basically a set of git hooks and scripts that enforce simple things AIs tend to skip—running tests, checking basic metrics, that sort of guardrail.

I think it’s useful for projects that started quickly, grew messy, and are now slowing down because the AI (and humans) can’t manage the complexity anymore. Take it for a spin and let me know what you think: https://github.com/CaliLuke/quality-workflow-meta

This is a fundamental misunderstanding of how LLMs work.

Local models can do embedding very well, which is useful for things like building a screenshot manager for example.

Do you have some kind of tooling to automate the process? Would like to try it.

I'm over simplifying here but graph database and knowledge graphs exist. An LLM doesn't need to preserve everything in context, just what it needs for that conversation.

Because that's not the job it was designed to do, and you would know by reading the article.

The article says it's not a good conversational model but can be used for data extraction and classification as two examples.

How many times do we need to hear "exercise and healthy eating" prolong life and improve cognitive function?

is there some sort of utility to automatically backup and retrieve dotfiles from github to keep different computers synchronized?

if a kid is being lazy there's simply no way around "cohercion" as you put it. You know how I know he's being lazy? Because I used to do the same stuff for the same reasons, and my parents and teachers saw through it and didn't make excuses for me or any other kid.

We were expected to grow up and learn to do work even when we didn't want to.

The crowd you see is because options are dwindling faster than people willing to use them.

As an immigrant to the USA teaching in this country is a mess. Teachers apply a lot of semi scientific mumbo jumbo to justify a completely inadequate amount of work required from students to learn.

I know it's not popular to say it but my son learns anything I teach him, he might not enjoy the process very much but he never forgot anything I taught him because I make him work. His teachers don't make him do anything with the results you can imagine. If you point it out they say if they did parents would complain.

but why?

and who wants that exactly?

that's not a technology problem, it's a political problem. You're trying to solve a political problem with technology?

I don't see anything obvious about the utility of cryptocurrencies.

Money is not that difficult of a concept, you use it to store value or transact. As a store of value it has no intrinsic advantage compared to any other alternative we have now, and same goes to transactions, it doesn't bring anything to the table than current systems already don't do.

All the features people keep mentioning like the immutable ledger are simply not something we actually need or government have asked for.

The only utility is making money for the people that work on it, there's no societal value in crypto at all.

I don't buy it, crypto doesn't bring anything to the table that improve either SWIFT or the gold reserve that we actually need, but I understand crypto product developers see an advantage in switching from something they don't make any money on to something they do. As a consumer, or in this case as a citizen, crypto presents no advantage.

maybe they don't know the gender and playing it like at parties when you don't remember someone's boyfriend name.

then it's not a TODO, it's just part of the comments. Promoting semantic diffusion IMHO is not a good practice in software development.

I rarely find reasons to disagree so much with anything I read like I did with this blog post. Seems like the author lacks rigor and then goes looking for ways to excuse being sloppy, that's an attitude I don't think is productive or creates anything positive in the industry.

it's not a mistake because it's what ends up happening. There's serious research on it that everyone seems to want to ignore: https://www.nber.org/papers/w23153