Open source software helped to dramatically reduce the cost of paid software, because there is a now a minimum bar of functionality you have to produce in order to sell software.
And, in many cases, you had to produce that value yourself. GPL licensing lawsuits ensured this.
AI extracting value from software in such a way that the creators no longer can take the small scraps they were willing to live on seems likely to change this dynamic.
I expect no-source-available software (including shareware) to proliferate again, to the detriment of open source.
> others who deem Tailwind valuable enough will continue to maintain it.
We have seen several examples in the last couple of years where this is simply not true enough. There are multiple open source projects that do not receive enough TLC.
If my company relies on an open source project and it isn't being maintained, I can either ask my company to start maintaining it or find something else or accept the risk of a an unmaintained project.
> LLMs are just a concept, an abstraction. A data type for storing data.
C'mon. You know good and well that what is being discussed is the _use_ of LLMs, with the concomitant heavy usage of CPU, storage, and bandwidth that the average user has no hope of matching.
> You know good and well that what is being discussed is the _use_ of LLMs
Not the person you're replying to, but I've found that some people do argue against LLMs themselves (as in, the tech, not just the usage). Specially in humanities/arts cycles which seem to have a stronger feeling of panic towards LLMs.
Clarifying which one you're talking about can save a lot of typing/talking some times.
> I've found that some people do argue against LLMs themselves (as in, the tech, not just the usage). Specially in humanities/arts cycles which seem to have a stronger feeling of panic towards LLMs.
Maybe?
The person I responded to said "LLMs are just a concept, an abstraction."
Were that true, were they simply words in some dusty CS textbook, it's hardly likely that the humanities/arts people you describe would even know about them.
No, it's the fact that these people have seen regurgitated pictures and words that makes it an issue.
> It's wild to me seeing the tech world veer into hyper-capitalism and IP protectionism.
Really? Where have you been the last 50 years?
> Plumbers are not claiming use fees from the pipes they installed a decade ago.
Plumbers also don't discount the job on the hopes they can sell more, or just go around installing random pipes in random locations hoping they can convince someone to pay them.
> Why should intellectual property authors be given extreme ownership over behavior then?
The position that cultural artifacts should enter into the commons sooner rather than later is not unreasonable by any means, but most software is not cultural, requires heavy maintenance for the duration of its life, and still is well past obsolescence, gone and forgotten, well before the time frame you are discussing.
You can even write down what you are looking at and listening to, although in some cases, dissemination of, e.g. verbatim copies in your writing could be considered copying.
But it is automatically copying if you use a copier.
What about a table saw with no guard, a wobbly blade, and directions from management to follow a faint line you'd have a hard time doing with a scroll saw?
Even if the LLM theoretically supported this, it's a big leap of faith to assume that all models on all their CPUs are always perfectly synced up, that there are never any silently slipstreamed fixes because someone figured out how to get the model to emit bad words or blueprints for a neutron bomb, etc.
Most of the cloud providers give you a choice of two ways of referring to models - either a specific dated model id (like the example above), or a shorter alias which generally points to the latest release of that model, and is more likely to change over time.
We add in some additional flags for `--opus`, `--sonnet`, `--haiku` as shortcuts to abstract this away even further if you want to just use the latest model releases.
Example to run haiku latest via Vercel AI Gateway with unified billing and cross-cloud fallback between providers.
`claude-run --haiku --vercel task.md`
AWS Bedrock at least appears to be pretty steady when you pin a model now, according to our own evals anyway. Earlier on there was some performance degradation, at peak load etc.
If galvanic isolation is necessary, there are "digital isolators" (that's a good search term if you are interested) that are much faster than optocouplers and that don't suffer from the same sort of degradation (over a few years, the LED gets dimmer and dimmer).
But there's probably no galvanic isolation going on here anyway, so a wire, or at most a simple logic buffer, would probably suffice.
If I'm connecting two things from different power domains, I like to use gates (or level shifters, if necessary) that are designed for the task. These will keep stray currents from causing electromigration problems when one is powered on and the other is powered off, and some of these are very fast, over 100 MB/s.
> over a few years, the LED gets dimmer and dimmer
That shouldn't happen unless the LED is driven near the top of its current rating, which shouldn't be necessary unless you're pushing the limits of its rise/fall times (in which case a different part would be advisable as you say).
A random app note shows 95% of initial current transfer ratio after 25 years at If = 5 mA, and depending on the necessary bit rate we could probably design for at least 2x initial margin on that CTR. Such a design would last effectively forever.
I think the galvanic isolation is mostly a feelgood here, allowing people to say it's "air-gapped" even though that's not directly relevant (since Wi-Fi is also "air-gapped"). A simple gate or level shifter can also enforce unidirectional data flow as you say.
> which shouldn't be necessary unless you're pushing the limits of its rise/fall times
Right, I should have clarified this. To make them go fast, you can often use more power (to a point), and that can shorten the LED lifespan. (To be fair, there are techniques to give you a bit faster speed without making them too bright, like pulse-shaping, but it didn't appear anything that fancy was going on there.)
And, unfortunately, "fast" for the optoisolator isn't very, in any case. The cut-off frequency for the first datasheet I found corresponding to that app note was 80 KHz.
> I think the galvanic isolation is mostly a feelgood here,
And...
I don't get this.
If it does nothing useful, why bother?
IMO, the primary good use for an optoisolator these days is either for something analog-ish like the feedback for a switch mode power supply, or for when you're breadboarding something with really high voltages and don't want to bother with SMT devices.
I think those optoisolators are indeed sold mostly for switching power supplies. That's probably why someone cared enough about aging to write an app note, since the ambient temperature is high there and the exact CTR matters more when it's in that analog feedback loop. I've also seen them for digital inputs in industrial control systems, where speeds are slow and the wires might be coming from far away on a noisy ground.
That said, I believe optical isolation is typical for these "data diode" applications, even between two computers in the same rack. I don't think it provides any security benefit, but it's cheap and customers expect it; so there's no commercial incentive to do anything else.
And, in many cases, you had to produce that value yourself. GPL licensing lawsuits ensured this.
AI extracting value from software in such a way that the creators no longer can take the small scraps they were willing to live on seems likely to change this dynamic.
I expect no-source-available software (including shareware) to proliferate again, to the detriment of open source.
reply