Go into planning mode and plan the overall refactor. Try to break the tasks down into things that you think will fit into a single context window.
For mid sized tasks and up, architecture absolutely has to be done up front in planning mode. You can ask it questions like "what are some alternatives?", "which approach is better?".
If it's producing spaghetti code, can you explain exactly what it's doing wrong? If you have an idea of what ideal solution should look like, it's not too difficult to guide the LLM to it.
In your prompt files, include bad and good examples. I have prompt files for API/interface design, comment writing, testing, etc. Some topics I split into multiple files like criteria for testing, testing conventions.
I've found the prompts where they go "you are a X engineer specializing in Y" don't really do much. You have to break things down into concrete instructions.
Does this support adding family members on a single account? I have some non technical family members who I'd like to manage it for them and giving them their own account is most likely going to be a major headache.
Hmm I’ve started doing some elixir as a side project and coming from typescript honestly I’ve found it a bit “dumb” and slow at the beginning. It didn’t allow you to write precise enough types, and would complain a lot about weird things.
As the project became more complex though, and I learned more about elixir, dialyzer became progressively more useful.
Most of the time when I chase a weird type error in TypeScript it turned out that the types were not correct enough and had to be fixed.
When I’m chasing weird dialyzer complaints, most of the time I discover that it was actually a bug in my code. Honestly sometimes I’m astonished how it can reason about very complex things so well. It’s just the error messages are usually very confusing, but with experience that kinda solves itself.
The difference to me so far has been that dialyzer finds bugs in my code, rather than in my types, which is quite nice.
If they make the built-in type system of elixir faster and with better error messages, I’ll be a very happy camper.
1. Since it's an extended display over wireless, bandwidth limitations mean you're limited to just a single 4k resolution monitor floating in space (they mentioned that it simulates a single 4k monitor in the demo)
2. You also would have to bring your laptop with you
To me, one of the major value propositions for something like the Vision Pro would be that it could replace your laptop entirely. But it doesn't seem like that will be the case, at least based on what they've announced.
I'm about 80% sure they'll come up with an USB-C -> Vision cable you can use to plug in to your Mac.
Or the power bank is an USB-C bank and you can just detach the cable from it and plug it in directly and use the laptop's battery + get a wired connection.
Well you can have the laptop as a single screen floating in space and then other floating screens that are running off the Vision Pro, e.g. your web browser, docs, etc.
Quest Pro weights 722g while MacBook Air 15 weights 1.5kg. How can you expect it's twice heavier than Quest? It's completely unacceptable for most people's head.
Hmm, I guess you could take it that way. But things that light only really get tiring if the weight is poorly distributed. It's more about the size and sort of bag it's likely to be in.
Not a professional but I think I follow markets pretty closely for a retail trader.
Twitter / FinTwit is the best
Level of discussion is not always great but I find its way better than anywhere else. Twitter Spaces usually have great discussions since people are less likely to troll vs tweets.
One thing I've found though is if the person is getting too cocky and going on victory laps, it's pretty much always a good time to counter trade them.
You'll be able to find some great Substack's, Podcasts, Youtube channels from Twitter as well if you're following the right people.
Some that I follow
- @fedguy12 - macro
- @JulianMI2 - macro
- @biancoresearch - macro
- @acrossthespread - Japan macro
- @anasalhajji - energy
- Pretty much any central bank, government official
- Fund managers. Many of them aren't too active and just crosspost links to videos. Cliff Asness is one that is active though
Podcasts:
- Forward Guidance
- Odd Lots
Reddit I find is usually not much better than SeekingAlpha, Zero Hedge, etc. even on the more "serious" subreddits. Most of the posters are John Bogle acolytes or post the Warren Buffet quote on buying index funds. A lot of the threads don't have much besides "you should buy a broad based index fund with a lump sum [link to Vanguard whitepaper] [link to backtest starting from year 2000-2010]".
HN is still way better than Reddit on the rare finance/economics post. Every finance thread you'll get a few people who have unique insights/opinions but I think the majority is still a lot of retail market participants who just regurgitate the Reddit stuff.
Yes it's possible ( https://www.ha-obsession.net/2017/05/u2f-sudo-fedora-25.html... and some other guides exists ) and I used it for about a year. I stopped using it. I was so annoyed to plugin the key every couple of hours that I simply kept the yubi key plugged in all the time - free to be picked by anyone and def. not increasing security.
Leave the yubikey plugged in all the time. It's fine with respect to most threat models, provided you lock the graphical session when you are away from the computer.
If someone steals the key, they can't really do anything with it. They can't sudo because the session is locked. They can't use it to log in your web accounts from other computers because websites ask for a password/pin in addition to touching the yubikey.
PS: you should always have a backup yubikey (or, better, two)
You can also set the yubikey to require a pin before touching. The yubikey auto wipes it’s memory of the presented pin is wrong too many times in a row. So just leaving it plugged in is much more sensible in that case.
I'm very curious to see how this plays out over the next year:
1. BTC dominance seems to be getting lower
2. Ethereum PoS seems very likely to launch this year IMO
3. Alt chains which run on PoS like AVAX and SOL have been eating into Ethereum's market share of the smart contract space. Newly onboarded retail users to crypto have been priced out of Ethereum due to fees over the past year. The alt chains running PoW seem to have no real usage.
4. Fed raising interest rates could dampen the growth of crypto asset prices
If I put myself in the shoes of a miner running GPU's, once the merge happens I only see these options:
1. I switch to a chain that is still profitable for GPU's. None of these have much demand for block space and unless that changes, other profit minded miners will be dumping the token.
2. Switch to some kind of P2P GPU-aaS like Render or Livepeer. I haven't been following this space to closely but I doubt it would have the same returns as mining Ethereum over the past years.
The criticism of catering to early adopters and whales is a bit weird to me. How is this different than any project that receives funding? What's the ideal solution for this?
Personally, I'm curious to see a project like Mastodon or Peertube fund itself using a token. I don't think those projects will ever get any meaningful use aside a few nerds and outcasts that got booted off of Web2 properties. You'll probably get speculators in it purely for profit, but to me it sounds better than the alternatives.
> Personally, I'm curious to see a project like Mastodon or Peertube fund itself using a token
Agreed. Most of the people criticizing Web3 are underestimating the power introducing token incentives will have. "Ordinary people" just don't care enough about decentralization of data for projects like Mastodon to take off otherwise. They've been around for a while, and they're not competing with Big Tech. A Web3 platform that rewards successful content creators natively with a token and not hearts/likes might though.
For mid sized tasks and up, architecture absolutely has to be done up front in planning mode. You can ask it questions like "what are some alternatives?", "which approach is better?".
If it's producing spaghetti code, can you explain exactly what it's doing wrong? If you have an idea of what ideal solution should look like, it's not too difficult to guide the LLM to it.
In your prompt files, include bad and good examples. I have prompt files for API/interface design, comment writing, testing, etc. Some topics I split into multiple files like criteria for testing, testing conventions.
I've found the prompts where they go "you are a X engineer specializing in Y" don't really do much. You have to break things down into concrete instructions.