< Back to the archive

Like what you see? Subscribe here and get it every week in your inbox!

Issue #228 - July 23, 2023

Here are the top threads of the week, happy reading!

Top comment by Analemma_

I don't remember where, but I read a lengthy comment from someone in the industry which said there were two main things preventing delivery drones from being a viable market:

1. FAA regulations - delivery drones can't operate within X miles of an airport (technically they can, but it requires a much stricter degree of certification and compliance nobody wants to bother with)

2. Drones need a landing space, so people without yards (like apartment and townhouse dwellers, who make up a lot of the population in exactly the densely populated areas where you'd want to use drones to begin with) can't be served

And it turns out that once you exclude "houses within X miles of an airport" and "houses without an LZ", there aren't enough customers left to make delivery drones worth it.

Top comment by thesuitonym

Good points have already been raised, but don't neglect the rise of smartphones. When you get a Google phone that already has Google's browser installed, and it works relatively well, why would you change? Add Chromebooks into the mix, and those meet a lot of people's needs, why change?

Then, if you get a laptop, you'll want to have all of your passwords and bookmarks synced, so instead of using Edge, you grab Chrome. never even thinking of Firefox.

Finally, in the early 00's, Firefox users had a reputation for letting you know about it. Forums of the day were full of signatures with a Get Firefox link in them. You don't really see that level of fervor anymore, because the difference between Firefox and Chrome today is nowhere near the difference between Firefox and IE back then.

Top comment by PaulRobinson

Claude Shannon's "A Mathematical Theory of Communication"[1] is often considered a classic. I think this is because:

1. It's quite readable as a narrative.

2. The maths is not pages of first principle derivations as if the reader is not familiar with the basics of algebraic substitution.

3. The diagrams and graphs are genuinely useful and remove the need for many, many thousands of words that others may have used instead of, or in addition to, the core narrative.

4. It deals with an abstract concept but roots it in concrete mathematical and physical terms. He touches on specific examples.

5. It's quite short given the breadth of subject area.

[1] https://people.math.harvard.edu/~ctm/home/text/others/shanno...

Top comment by tmikaeld

The most low-cost I know of with widest reach (most countries) is https://www.adyen.com/pricing

Because you can choose which processor you want to use and there are many low-cost ones, including some inter-bank ones with fixed cost (no %)

Most shops like WooCommerce and Shopify have ready-to-use plugins for it.

(I'm not affiliated, but i build e-commerce for brands)

Top comment by jrajav

React is now the de facto choice and much less painful than past de facto choices, is my take on why. It has a mature and healthy ecosystem, and its relatively small surface makes it easy to learn and predictable, while declarative GUI in general is low on footguns compared to ye ol two-way data binding.

Many of the competitors springing up define themselves in terms of React, either as an improvement or a foil. But if you're just trying to get things done, it's hard to go wrong with it - and in a practical sense, not a "nobody got fired" sense.

Top comment by ramoz

At a large government agency, partnered with one of the top AI companies in the world. We've been deploying models and transformers and recently a LLM ensemble in prod for 3-4 years now. Many lessons learnt that open source hadn't really provided utility for.

The biggest backend things: models can run on CPU architecture, you still need highly performant CPU, but in cloud there is a way to significantly discount that by relying on spot instances (this doesn't mean GPU isnt viable, we just found a real cost lever here that CPU has supported... but things may change and we know how to test that). Further, distributed and parallel (network) processing are important, especially in retrieval-augmented architectures... thus we've booted python long ago, and the lower level apis aren't simply serializing standard json-- think protobuffer land (tensorflow serving offers inspiration). BTW we never needed any "vector DB" ... the integration of real data is complex, embeddings get created through platforms like Airflow, metadata in a document store, and made available for fast retrieval on low-gravity disk (e.g. built on top of something like rocks db... then something like ANN is easy on tools like FAISS).

The biggest UX thing: integration into real-life workflows has to be worked on in a design-centered way, very carefully. Responsible AI means not blindly throwing these type of things into critical workflows. This guides what I assume to be a bit of a complicated frontend effort for most (was for us esp as we integrate into real existing applications... we started prototyping with a chrome extension).

EDIT: btw oss has definitely made some of these things easier (esp for greenfield & simple product architecture), but I'm skeptical of the new popular kids on the block that are vc-funded & building abstractions you'll likely need to break through across software, data, and inference serving for anything in scaled+integrated commercial enterprise.

EDIT2: Monitoring, metrics, and analytics architecture were big efforts as well. Ask yourself how do you measure "value added by AI"

Top comment by bmitc

The concepts in SICP are timeless. There's nothing to be lost going through that book.

There are video lectures you can use to accompany the book: https://youtube.com/playlist?list=PLE18841CABEA24090

Also, I highly recommend the courses How to Code: Simple Data amd How to Code: Complex Data on edX that are based on HtDP.

Top comment by fidotron

Capital 'A' Agile has become a total anti-pattern, and the most revealing signs of this are the two-week sprint, excessive meetings, and lack of technical leadership.

By far the most productive teams I've worked with have been on six week iterations, with an at least approximate idea of what they're supposed to achieve in the next 90 days. One week out of that six is basically given over to demo, retro, and working out what the next one should achieve, but there is enough slack in there to accommodate the inevitable ups and downs, which two week Agile/Scrum pretends is not a thing. The lead developer is almost entirely focused on integrating the work of the others, and building strategy for the next iteration with product/marketing and the other team leads.

The other incredible tendency of highly productive teams is a lack of communication . . . because they don't need it. It may be more accurate to class this as a lack of noise, and clarity over what needs delivering. (i.e. product and the leads have done their job). I used to have to visit different teams in many locations, and you could tell how productive they were just by how quiet it was, interrupted only with the occasional "Dave, you broke the build!"

Top comment by nicbou

My movie triaging and conversion thing, basically a personal Netflix. This one has been rock solid and requires zero maintenance. There are better torrent-based options nowadays, I think.

https://nicolasbouliane.com/projects/home-server

The other is my timeline thing. I love the idea, but messed up the implementation. It's a bloated mess that I dread working on. It should be a presentation layer for my /backup directory, not an opinionated backup solution cum data archive management system.

I plan to rewrite it into a much simpler project.

https://nicolasbouliane.com/projects/timeline

The last one is Syncthing, since a week or two. It's much simpler than Google Drive or the timeline thing's elaborate rsync setup, so it will replace both.

A while ago, I read that "a person's main task is not computing, but being human". This has heavily influenced my relationship with technology, including the tech I create for myself. I want calm technology that blends into my life, and self-hosted software is rarely that.

https://calmtech.com/

Top comment by pezo1919

Unreal Tournament (99 / GOTY Edition)

I was ~6 years old, when I played the demo version. Earlier, I had played Q2, (maybe Q3 demo?!), Tiberian Sun, Red Alert before and couple of simpler games and demos.

Unreal Tournament demo had like ~4-5 Deathmatch/Team Deathmatch maps, 1 Capture the Flag (infinite gameplay for me at that age on CTF-Face (Facing Worlds) :D ), maybe 1 Assault map, but not sure.

The motor, feeling, control, sounds, music were just pure awesome. Sniper headshots, jumps, tricks, extremely competitive feeling, double kill, triple kill, ultrakill, mo-mo-mo-mo-monsterkill... :D

Then the Game of The Years version was a Christmas present for me, I still remember the smell and the feeling. I could not try it out for a bit, so I was reading the (Hungarian! :D) manual back and forth. The GOTY version had extra weapons, maps, skins, goodies.

Later, internet was slow, but mods and couple servers were just awesome.

I want to go back! :D

Unfortunately it is impossible. Partially due to the missing scene, partially because I get frustrated, when I realize that my muscle memory (aiming) is totally broken and I cant hit shit, even if I truly see and feel, that I hit. :D

Back then, I did not even see if I hit, I just knew. And I truly hit. Wtf, it's gone. :D

Nowdays the only maps I play are TS/Expo/mobx-keystone. :)