30,161 karma · joined March 17, 2008
Working on sync engines, PGlite, and Tanstack DB at Neon/Databricks. Previously co-founder ElectricSQL (acquired by Databricks).
sam dot willis at gmail dot com
samwillis.co.uk
Twitter: @samwillis
The centre of the signal was something like 69 million KM off to the side of earth. 20% of of earths orbital diameter.
I suppose the difference is the effective receive angle, and by turning up the power they could get it to respond.
As the files are just on the file system, what's the sync story? Have you tackled merging conflicting edits across devices?
GVR for example works for Microsoft and has worked at Dropbox and other places, but spends much of his time on Python. So he isn't listed as an employee of the foundation, but at times is full time.
If a company donates employee time they have more influence on the project than donating cash to pay for development.
The fact that Zig had the cash donation to do it this way is brilliant, and probably a better model.
Last year I had a month free and I had a go at building something for Django in this area, with a bunch of interesting ideas - built on Alpine.js, resumable server side component state, inline component templates. But sadly time is limited and I just can't spend the time needed to push it further. One day I may be able to pivot back to it, but would be very happy for someone to take some of these ideas and run with them - https://www.tetraframework.com/
What I really want to see is a modern asset + front end build solution for Django. It would enable more of these sorts of projects. Thats what needs to be built into Django.
Quite right CRDTs are not magic, but they do get you something like 80-90% of the way there for most CRUD style apps. There is also a bunch of work going on into developing new CRDTs for different use cases.
What's important is articles like this help us to explain to less technical people what local first is, the advantages it has. They help to sell the concept.
But ultimately Wired is entertainment isn't it.
There are many ways to build local first, both with web technologies and with "native" app toolkits.
Many of the CRDT tools have started as JS/TS libs but have been rebuilt in Rust to be cross platform, WASM or native.
The appeal of using web tech is its accessibility and ease of distribution. Local first and PWAs with the new origin privet file system api kind of go hand in hand, but it's not an exclusive relationship.
Somewhat counter to the article, I would argue that local first isn't about having "no cloud", it's "low cloud". So you could go P2P for sync, but you can also have a much more lightweight sync service on a central server. Both are valid "local first", it comes down to use case.
The way to think of the term is like when we all moved to designing web UI "mobile first". It's a kind of shift from operating one way and adding on support for the other, to the other way around. Start local and layer on server and cloud support.
One of the most exciting things about local first though is that once you have solved all the sync, and conflicting edit resolution problems (CRDTs are key here) you are 90% of the way to solving "multiplayer real-time". The only missing bit presence.
The community around local first is also really buzzy, it's got that great feeling of being part of something new, something that's going to be bigger than the sum of its parts.
To learn more about local first this is a great starting point https://localfirstweb.dev/
Within Search + Search Ads you have a platform where Google are/have:
- the sole vendor of a digital asset - ads
- the sole marketplace for that asset
- a closed algorithmic trading system with no public auditing
- no public ledger or auditing showing who bid what
- currently one of the main customers for those ads (advertising their own products)
Along with that you as an advertiser have to install tracking on your site allowing Google to see every transaction, all revenue, every customer, of your business. The level of spying on the advertisers is insane, when seen in combination with the closed source algorithm and market for the ads the potential for manipulation is enormous.
The whole ads platform is built for extracting every last cent of margin from advertisers businesses. It's anticompetitive, monopolistic, and to some extent even a protection racket.
The most important "breakup" needs to be the ads business from everything else, including search. That's going to hurt, and I don't have the answers to how to make it work, but changes are needed.
I suspect a "breakup" isn't on the cards, maybe some strong regulation of online ad marketplaces, ring fencing them from the rest of the business and auditing of the platform?
if a.b.c:
print("c:", a.b.c)If someone pays with a card, then disputes that they did not receive the service or product, the burden is on the vendor to prove they did. It's not the card processor who makes the final call either, it's the customers card company - it's out of Stripes hands.
We got the crossover when we stated a family, it felt big and safe (compared to what I had before). However, the boot/trunk was too small for a growing family going on trips.
We upgraded to a large SUV, felt great to drive, big a tall. Masses of space to load up, however hard to park in small spaces (this is the UK), and far too expensive to run.
Now on an estate/station waggon, love it, favourite car I have owned. Cheaper to purchase than equivalent size suv as they are less fashionable, plus cheaper to run, cheaper tires for example. Far more fun to drive than both the others (and anything else I've had). Do not miss the height at all.
Sure local inference if harder than in the cloud, but hardware is getting better all the time, and we are still early on the optimisation curve when it comes to LLMs.
I'm looking forward to hopefully seeing smaller less resource intensive models that are easer to run locally, even on mobile.
Does anyone who of any research info "trimming" or "stripping" less used parts of a LLM so that you can take a trained weights and make them smaller (obviously with loss of some sort)?
The public at large see a rich person with an opportunity to grow their wealth through politics. You may well be right, and that is the (original) concept behind the House of Lords, but as with everything in politics its appearance that matters.
Also it's not always about them being corrupted, it can be them as the corrupting force.
The company, Infosys is his father in laws (his wife does have shares), and as far as I can see there is no link between this IT contract and North Sea oil rights. But maybe I'm wrong and it a simple bribe, but I doubt it.
Fundamentally though, a billionaire should not be a political leader, too much opportunity of corruption or the appearance of it.
The point is, by the time you have solved all the problems with consultancy and sync to an edge node, you are most of the way to local first.
Suhails postmortem: https://twitter.com/suhail/status/1591813110230568963
The OP immediately made me think of Mighty.
PUT - is effectively an "upsert" at a specific url. Doesn't exist? Create it, does exist? replace it.
PATCH - update a resource with a diff, at a specific url.
POST - this is a RPC, in the case of a REST API it can be used to create a new resource where the "id" is not provided and set by the server, it then redirects to the new url.
POST can be used for any RPC endpoints, even as part of a REST api.
ICE cars all show a percentage, and maybe additionally a mileage on newer cars. The fact that they threw that away is a little silly.
However, I also dislike so much of it. The hacks I have used to contort it to do what I want I would be embarrassed for another developer to see...
Former Django fellow Carlton is working on a nice toolkit called "Neapolitan", it's early stage, but is designed to enable building CRUD views very quickly. There is a good chance that in time it, or a layer on top of it, will become the perfect successor to the Admin.
Give users control of their data and privacy, moving away from large backend infrastructure, and providing better value to the user. With the bonus of having a more enjoyable development environment and an excited community.
Show HN: PRQL 0.2 – a better SQL - 378 points by maximilianroos on June 27, 2022, 161 comments
Original post when it was conceived: https://news.ycombinator.com/item?id=30060784
PRQL – A proposal for a better SQL - 650 points by maximilianroos on Jan 24, 2022, 295 comments
HN guidelines: https://news.ycombinator.com/newsguidelines.html
> Please don't complain about tangential annoyances—e.g. article or website formats, name collisions, or back-button breakage. They're too common to be interesting.
The devs will certainly have a reason for having a captcha, I suspect it keeps the actual game fun rather than being overrun with bots.