Probably a good place to post a 400 year old recipe for pancakes. I've made them, they are very good. Note that what is considered a "pancake" has changed over the years and changes with location.
Off topic, but just in case: is there a good reference on how people actually use LLMs on a daily basis ? All my attempts so far have been pretty underwhelming:
* when I use chatbots as search engines, I'm very quickly disappointed by obvious hallucinations
* I ended up disabling github copilot because it was just "auto-complete on steroids" at best, and "auto-complete on mushrooms" at worst
* I rarely have use cases where I have to "generate a plausible page of text that statistically looks like the internet" - usually, when I have to write about something, it's to put information that's in my head into other people head
* I'd love to have something that reads all my codebase and draws graphs, explain how things work, etc... But I tried aider/ollama, etc.. and nothing even starts making sense (is that an avenue to persevere in, though ?)
* At once, I tried to write in plain english a situation where a team has to do X tasks, in Y weeks, and I needed a table of who should be working on what for each week. I was impressed that LLMs were able to produce a table - the slight problem was that, of course, the table was completely wrong. Again, is it just bad prompting ?
It's an interesting problem when you don't know if you're just having a solution in search of a problem, or if you're missing something obvious about how to use a tool.
Also, all introductory texts about LLMs go into many details about how they're made (NNs and transformers and large corpuses and lots of electricity etc...) but "what you can do with it" looks like toy examples / simply not what I do."
So, what is the "start from here" about what it can really do ?
> I'd rather get to value now by making something that just works(and is adequately tested) than engineer something thats future proof but takes longer to get out.
If you haven't read it, I highly recommend A Philosophy of Software Design by John Ousterhout. His talk is a great teaser for the book: https://www.youtube.com/watch?v=bmSAYlu0NcY
Here's the thing: a program properly written for the future does not take longer to get out. Not much, at least. What we instinctively think of as "future proof" is generally anything but: by anticipating some particular kind of change that may happen some time in the future, but in practice almost never does, we end up with more complex programs that are harder to modify when an actual (typically unforeseen) change comes our way.
The best way to future proof your program is to make it simpler, according to the requirement you are currently aware of. If you expect a particular change, sure, plan for it. But for the unexpected, just keep it simple. Because simple is easier to adapt, add to, or rewrite.
That said, the simplest solution is rarely the most obvious. Making thing simple does take a bit more time than rushing through the first thing that comes to mind. But it also pays off in a matter of weeks, so it's quite worth it.
And if you have a hard time assessing what "simple" means, SLoC is a surprisingly good metric (we have research that shows this). And mostly your modules should be deep: small APIs that hide significant implementations. I think of it as encapsulation on steroids.
React Native, which was meant for writing amazing native apps, is just misused as a browser web-view thing. I am an expert in this field. Just never use React native to open single WebView. Never try to turn your website into mobile app like this. If it's a website, it's a website. If you need app make app. React, Electron and React Native enables you to make apps using Javascript, which is same as your website, it means you can have the same business logic between web and native app, same language it all written in, team talking on the same language, only the rendering logic will be different, which is huge saves. But it's not Flutter. It was not meant just to present webviews with your webpage. There is a Browser for that. Finally there is a PWA for just exactly that.
Amazing stuff which React + Electron + React Native enabled - abstracting JS from Web to App - is so much crooked by folks trying to turn it inside out
React Native is not the future – for folks who don't understand how to use it. In fact they hate using it wrong.
For me – it's the future, i can use only one language to create an Android, iOS, Windows, Linux and Mac native apps. (of course, i would never try to fit it into a single front-end codebase, this is pure madness)
It seems to me that many of the current changes in society are not really changes, but just the ending of what was an abnormal little, extremely pleasant, bubble in time. And it's in that bubble that "we" (Westerner, over the age of ~30) all grew up in.
"Nothing can now be believed which is seen in a newspaper. Truth itself becomes suspicious by being put into that polluted vehicle. The real extent of this state of misinformation is known only to those who are in situations to confront facts within their knowledge with the lies of the day. .... I will add, that the man who never looks into a newspaper is better informed than he who reads them; inasmuch as he who knows nothing is nearer to truth than he whose mind is filled with falsehoods & errors. He who reads nothing will still learn the great facts, and the details are all false." - Thomas Jefferson, 1807 [1]
That sort of condemnation of the news would have been, at the minimum, quite hyperbolic during the time we grew up in. But now? It seems like a scathing, yet completely reasonable, criticism of the current state of media. And this goes on endlessly far back. Read the classics (Plato's "The Republic" feels effectively prophetic at this point) and it goes from sounding alien, during our bubble, to sounding like something oddly familiar.
It's quite comforting in a way. We're not stepping into some great new unknown, but merely returning to the endless cycling of history.
I would like to offer some tips for those faced with the potential of layoffs that I have compiled. I understand much of these come too late for those already affected, but for those worried about the prospect, these can help to ease the pain if it does happen:
• Check if your company pays out unused PTO, sick days, etc as cash. If they do, do not use any of the applicable type(s) unless you are going to lose it.
• Have a LinkedIn, fill out all the fields, add 500+ random people in your field. Once you have done all this, you get ranked way higher in the algo for recruiters who are searching (you will be granted a visible "All-Star" status, so you will know when you've reached this). After that, go add every recruiter in your field/industry you can find (ideally 500+). Internal recruiters are better than external recruiters / headhunters, but don't neglect the headhunters, especially the "rockstar" ones from more prestigious staffing firms. Finally, add a bunch (500+) of people in your field (who you should now have mutuals with, via the recruiters). Always respond politely to all recruiters even if you're happily employed. Try to be friendly with them, not strictly professional. Build up a rolodex of recruiters. You now have a list of people you can ask for work if you do get laid off. Recruiter-sourced candidates have MUCH better odds of being hired than cold applicants, provided you're not a known name in your industry. If you do this, you'll be able to schedule 40+ interviews in about 3 days, which take place over the following week or two, if you really want to pack them together.
• Don't neglect contract work completely. Many companies have a surprisingly large hiring pipeline of contract -> FTE, provided you do a good job.
• How To Win Friends And Influence People by Dale Carnegie.
• Corporate Confidential by Cynthia Shapiro, if you're in an enterprise / corporate environment.
It's not accepted as fact, but an interesting theory on "WTF happened in 1971" was the "Legislative Reorganization Act of 1970" and how it made many actions, votes, etc, of the US congress more transparent.
The idea being that lobbyists were paying for certain types of votes when they were private, but that politicians could then vote their conscience anyway...because there was no way for a lobbyist to verify they got the vote they paid for.
But, once the specific vote by named representative became public record, the lobbyists could now see if they got the vote they paid for.
I really loved Shenzhen I/O, it even got me into electronics. I started watching Ben Eater’s series on breadboard computers and built some similar devices to what I was making in the game around a 6502. It’s pretty cool that a game can trigger something like that.
The game had a leaderboard, and at one point an old programmer colleague reached out to me to ask if I was cheating - that’s something I’m still very proud of to this day!
Users can interactively explore the microgpt pipeline end to end, from tokenization until inference.
[1] English GPT lab:
https://ko-microgpt.vercel.app/