Combine that most CS students learn many languages with LLMs and coding agents and the size of the ecosystem isn't quite as important as it used to be. New hires can be productive from day 1. Missing libraries are relatively easy to add. Moreover the language characteristics can be more useful than ever: fast running, fast compiling, typed, easy to read, etc.
Yeah I think LLMs really help with the chicken-egg situation in language adoption. Contrary to many opinions that predict programming homogenizing around the big 3 languages that exist today (because that's what the LLMs currently write) I think in the future more nice languages will gain adoption as they are written by LLMs, who as you note don't care about a lack of community surrounding those langs -- if they need a missing library the AI can just write it. Maybe they even add it to the language ecosystem for other AI or humans.
I think Python is actually kind of the worst language of the top langs to be the lingua franca of AI, where more niche statically typed languages like Nim are better suited.
As a Pythonista I tend to agree. I had high hopes for Mojo but it's taking its due time to become usable outside the narrow focus of GPU programming, whereas Nim fits multiple niches surprisingly well.
One of my concern is LLMs are going to generate a lot of low quality code for languages that do not have sufficient discussions on forums like Stackoverflow.
That's why these niche languages need state-of-the-art compilers that enforce invariants more strongly. This way, they can catch most of the subtle bugs the LLM produces, sort of like antibodies.
I should give a talk. About what however? I’ve been happy with my progress on FigDraw (1), a 2D UI scene renderer using SDFs. Even made my own neovim ui shell with it!
It’s been true for a while, but AI seems to have exaggerated it. To me it reinforces the idea that LLMs created a “K” shaped talent market. Those whose are good become even more valuable.
It makes me think tech communities need to lobby for more laws to ensure fair access to platforms, app stores, etc. Be that at least side loading apps, etc.
Otherwise we’ll eventually all get lost in the kafkaesque technocracies.
Less for moral reading, but to keep from being squashed by the weight of tech.
But eff isn’t going to come to my aid if it’s isn’t a big story, like wireguard. We’re all just arguing circularly around the fact that companies with massive footprints can and do operate in a manner where it’s assumed that zero access is the industry standard for “normal users”
At the same time the dumps of government and corporate emails have been invaluable to society at large. They’ve helped win court cases, uncover corruption, etc.
Nah, ESP32's have had ethernet capability for a while and ESP-IDF supports it well. I've been using one I built for 5+ years now. Unfortunately RMII (ethernet phy) interface takes up a lot of the GPIO pins. This part looks like it'll remedy that issue.
There's two ESP32 boards that have been around for a while with PoE:
> I'm more hopeful for single-pair ethernet to gain momentum though!
I keep looking for a reasonably priced 10baseT to 10Base-T1L bridge... everything commercial seems too expensive (for me) and the two hobby designs [1] [2] I've seen are not orderable :(
But I'm seeing more commercial options lately, so that's hopeful.
Multidrop SPE isn't going to outperform newer CAN versions though. Somewhere in the sub-100Mb/s (e.g. 10-20Mb/s range) is the practical maximum speed of a multidrop bus at useful lengths, and that essentially applies equally to CAN or SPE. The only way to really get faster in a "multidrop-like" sense is with logically loop-like systems like ethercat and Fibre Channel where each network segment is point-to-point and the nodes are responsible for the routing.
Can't wait until AI companies go from mimicking human thought to figuring how to licensing those thoughts. ;)
reply