I am not sure why this old news is surfacing here today but I can give my 2 cents, since I sold speedchecker.com last year and were directly competing with Ookla.
The main business is selling the data. You use Speedtest.net to troubleshoot your connection but metrics captured with the test alongside location data give telcos invaluable insights on where they should improve their networks. Telcos pay 6 figures annually for this data and we have a few hundreds of of those big MNOs globally. This market is pretty big. Accenture is in trouble with their main consulting business due to AI so acquiring data business is one of the smart strategies they can implement to stay relevant.
To all commenters who think they can code it over the weekend, yes you are right. I coded my first speed checker over the weekend in 2008 but it took me 18 years to grow the user base , figure out entreprise sales strategy and exit. Its not easy as it seems.
If you are politically connected, or stay in an narrow lane of approved work, you get your grant. But if you stray from the politically approved path, or appear disloyal to our First Citizen and the Party, then your grant will be canceled.
The remaining supporters of the incumbent party like to claim that they aren't actually doing anything worse than in the past, and if anything they are just cracking down on things that they see as subjectively bad, so it's fine. And there's an element of truth in that: so much of American policy for a long time has been subject to agency interpretation and judicial review, and there was always room for political maneuvering and corruption in the system. Where the truth becomes a lie is the omission that this is the systematic ramping up from something that happens occasionally in a mostly-functioning system, to something that happens constantly and is systematically designed to facilitate corruption and politicization.
While I agree that domain expertise has always been a moat, I believe the author is missing something critical: there is a big difference between being able to verify the output of a system is correct, and being able to tell a system how to generate the correct output to begin with.
Personal example: I had a software engineering colleague who was the best coder of financial management systems I've ever encountered. He gained these skills through years of in-the-trenches development. One of the things he told me, and that I also observed, was that the vast majority of financial experts (basically, the people in the accounting department of companies) had an extremely difficult time just telling him what the rules of any particular transaction should be. But what they could do was tell him whether the handling of any particular transaction was right or wrong. So often times he would sit down with these accounting folks and go through lots of example transactions he came up with, and from there he essentially built up the requirements spec.
In my experience, that is the primary difference between people I've known who are good software engineers and those who aren't: people who can specify the detailed rules of any system, vs. folks who take a "well, I know it when I see it" approach.
I have a strong suspicion that folks who have a high degree of domain expertise in a particular area will fail as software builders even in an agentic world because they will struggle to elucidate clearly the rules in their head that they've learned over years. As an analogy, it's kind of like asking a native speaker for the grammar rules of their language. Often times they can't, but they'll just say "well, that sounds wrong." They may be "domain experts" in their language, but they'd have a hell of a time prompting an AI system on how to grade a test for grammar correctness.
I never want to hear from developers again that they are not susceptible to marketing. I see meet ups specifically about Claude often.
Modern tupperware party.
A colleague was convinced Claude is better so we played a game. We used the claude code and codex harness and I implemented some prs they needed with gpt5.5 and opus4.7 and asked them to identify which came from which only from the code.
Couldn’t tell.
Edit: i bet 99% of people here, if presented with a test where i gave 5 models but all of the results came from one, would not be able to discern this. Just vibes all the way down.
It took me quite a while to come round to OpenRouter. Originally I didn't understand why anyone would put a proxy between them and an LLM, but it actually adds some quite significant value:
1. By far the lowest friction way to support and try out all the models.
2. They offer billing caps! Most model providers still don't do this [EDIT: maybe they do, see reply comment], but if you're going to run anything in public it's very useful to have hard limits so it doesn't cost you $1m overnight because someone started abusing it.
3. Their rankings are one of the more interesting signals for which models are popular, despite their flaws (most OpenAI and Anthropic users don't go via OpenRouter, it's currently not possible to tell the difference between many users switching v.s. one "whale" changing their preferred model)
Given how API costs are becoming meaningful for a lot of companies now, having a provider like OpenRouter to help measure your spend and easily experiment with and switch providers feels like a valuable service.
This is the new way and we need to stop it now. Forget the 'is it legal or not' arguments, their lawyers will win. Just get mad and tell them this is wrong. Stop buying their #$@#$ software. Block them. This is what is wrong with cars too. Don't want to give them real time data on you and your passengers and instead try to disconnect the modem? Well, no car functionality for you even if it doesn't need it. -get mad- Stop taking it. Microsoft is the enemy and needs to be treated that way. Same with any tech company that does the bait and switch TOS world. I buy so little software now and it is hard, but unless we stop this now it will only get worse.
> If AI suddenly makes it possible for a law firm to be run with a skeleton crew, then what's stopping all those people you fired from starting new law companies, where AI also does most of the work, and competing with you for the same market?
Money. They won't have the money to pay for the tokens, or the best models, because they'll be unemployed. They also won't have the connections to get the clients.
When you're playing a game of "who has the best capital," the scrappy underdog worker with vastly less won't win.
The idea that making the economy even more capital intensive will some how equalize things is an insane fantasy only a software engineer could swallow.
"Agent Readiness" will likely age as well as "Web 4.0 Blockchain Integration" has.
(To be entirely clear, not because agents won't be a relevant thing, although certainly I have my doubts, but because I believe even if they are a relevant thing, requiring special allowances from sites undermines the whole point, and such things will only end up used by bad actors to mismatch what agents see to what humans see, and so will be intentionally ignored.)
It has become a meme at this point but this sentence still stands: "The underlying purpose of AI is to allow wealth to access skill while removing from the skilled the ability to access wealth".
I think Sam Altman is an asshole and I prefer to spend my money elsewhere.
Frontier models being commoditize is inevitable. OpenAI thinks they're still competing on technology, and not user experience and market reputation otherwise they'd understand the continuous negative PR generated by Altman's chaos is going to cost them everything.
This change would go against multiple consumer guarantees in Australia where it's 1) a right to have undisturbed possession of a product 2) products must be fit for the advertised purpose https://www.accc.gov.au/consumers/buying-products-and-servic... Microsoft would be breaking consumer law if the change goes ahead for the perpetual licenses they sold in Australia
Cloudflare is known to use fingerprinting to detect scrapers For example, they use JA3 fingerprints and match them against the UA to block stuff like cURL while allowing OkHttp (Android clients) - but this can be easily be spoofed with packages such as CycleTLS [1].
I don't want to defend them, because they gate away a good chunk of the internet with their "bot protection", but unless you do PoW (which is also ecologically a nightmare), probably fingerprinting is the way to go - completely destroying the privacy of everyone involved.
Cromite, a privacy conscious fork of Chromium for Android, has constantly issues with CloudFlare Turnstile [2] because they (Cloudflare) try to fingerprint it in multiple ways in order to pass the challenge. The only way to get it to work would be to join the CloudFlare Browser Developer program - which requires signing an NDA. Rightfully so, the project maintainer didn't want to do it.
If you want to see the extent of what CloudFlare does to fingerprint the browsers, just have a look in the issue [2] and see which flags need to be disabled in order to allow CloudFlare to pass the challenge.
I understand both sides, but at least CloudFlare could be flexible enough to fall back to PoW instead of just blocking people from sending forms or accessing websites...
How much pontificating needs to be done before people acknowledge nobody has any idea what to do with AI on an individual level?
First being good developer and learning how to use AI was sufficient, next it was being able to design architecture, then it was “taste” that made all the difference and now being an expert in the domain is the only thing that matters really.
Until AI is basically in a stable, predictable, state of improvement or stagnation, these takes will continue to be pointless and most likely completely wrong.
The abrupt swing in many non-technology company IT departments from "hey developer, you aren't using enough tokens" to this is just too funny.
And I'm seeing almost no self-awareness from leaders. They are making decisions about things that they just don't understand. And are completely unworried about it. Just blindly following whatever the news cycle is about AI.
You have a rock solid piece of software used by an infinite amount of people and other services. It works fine, does it's job and just have some time to time updates due to minor bug fixes.
Why do we need AI here?
And more over, why people is saying "fork it and use the previous version". It should be actually all the way around, create a parallel fork younamethetool-ai and keep the OG untouched.
What I have to do now, keep a fork of my entire system's toolkit?
I wonder how many of the responses here bifurcate by age. The post resonates with me, but I am now in my early fifties. When I was in my 20's and 30's, I would have happily chased rabbits down all those holes, but now that time seems so brutally finite, I feel that anything encouring me to spend time on stuff other than what really matters is a strong negative. (Where "what matters" includes work, family, friends, and recreation).
When friends start dying within 10 years of your age, it's a hell of a wake up.
"I wish I'd made more throw away apps I never use" ... said no one on their death bed, ever.
The main reason is that it adds another layer (and human) that can, and probably will, get out of sync with the real-world implementation, whether that implementation is an API, web, or a CLI.
AI should not be using a protocol or set of instructions that is different from what humans have access to (know and use).
Sure, companies want to expose MCP servers because it is the cool thing to do right now.
So the current situation is basically that I used Claude to write an MCP server on top of our API. And then I need to occasionally tell it update it match the public doc.
And my reaction is: really? It is not like our API docs are not public. Claude Code created our MCP server with zero instructions beyond what is publicly available. I just told it to read the docs from the net.
So MCP feels more like a temporary workaround for current model limitations.
I swear to God. I just want to go back to the 2000s where everything was just plain HTML and some basic CSS, if at all any, by default you got responsive design out of the box, readable text and super user friendly GUI from the browser's own default stylesheet.
Today you open any website. Everything is a fucking component. A simple dropdown with a finite list? Has its own loader and makes 10 fetch requests for no reason. Not even exaggerating - look at Instagram and Facebook on web.
Fuck all these specifications, just give me the raw HTML that isn't obfuscated by your shitty/shiny new JS framework that you swear will change the game (looking at you, React)
> We used the claude code and codex harness and I implemented some prs they needed with gpt5.5 and opus4.7 and asked them to identify which came from which only from the code.
> Couldn’t tell.
Why would you expect them to be able to recognize the signature of a model from a pair of PRs? I don’t understand why you think this is a useful test for anything when we have numerous benchmarks that run 100s of tests on models and both GPT-5.5 and Opus-4.8 perform similarly.
I have subscriptions to both. I run both on max reasoning. It is interesting to see the relative strengths and weaknesses of each model. You won’t always see it if you’re just scanning code. Some times one will spin for a long time on certain problems where the other has no problem finding the appropriate parts of the codebase and getting an efficient solution.
antirez made a comment that he and others found GPT-5.5 to be better at the optimization tasks he was working on than Opus. There are other classes of tasks where GPT-5.5 consistently stumbles where Opus will get a solution quicker. Lately I’ve been working on some code where neither model comes up with a good solution. That’s just how LLMs go.
The only reason you have seen more activity about Claude is that they got there first. Codex has been a step behind and GPT couldn’t match Opus at first. You’re testing them after they’ve closed the gap.
I find the way that issue was opened incredible obnoxious, but it is baffling that the maintainers seem to have let AI loose on rsync. Like, why? Why try comparatively experimental crap when your fortune and reputation is made and you're the leader of a niche and immune to market pressure and the people love the thing and it does exactly what it's supposed to and works well?
It's like the Matrix, with the little rant about the primitive human minds not being able to accept paradise. You wrote the perfect tool, you won, almost undisplaceable in a niche, reliable, a metaphorical household name. It makes no sense to anyone to gamble or mess with that, it's just mind boggling.
And that's still a damn obnoxious thing to do in the formal issue tracker. Bad attitude, bad faith.
This whole brigading is bizzarre and some people are behaving like irrational animals. I potentially understand the motivations that might bring one to want to "win" this battle but this really isn't it - it just makes you sound like a fanatic.
It takes 5 minutes to search for "regression" on the issue page and go through the 17 results. There are potentially even more on the tracker used prior to github.
I think this behavior is very silly and people are just trying to justify their hate to AI by latching onto every possible thing, seemingly forgetting that before AI people did mistakes as well.
If you have proof that AI involvement in rsync has lead to a significant increase in open issues please show it to me - I'll be happy to change my mind.
The way this position conforms to the interests of the capital class, and conflicts with those of the labor class, is a red flag.
It simultaneously and conveniently:
1. takes the heat off AI blowback
2. synergizes perfectly with "RTO" mandates (to the extent this needed synergy to become A Widespread Thing)
On that basis alone, I'll wait for further analysis.
Edit: to be clear I'm no anti-AI holdout, and I actually don't mind working from the office (which i do 4x a week). Just observing.
> I don't want to defend them, because they gate away a good chunk of the internet with their "bot protection"
They also gate away a good many people with their "bot protection". I am extremely worried about how so many seem to have outsourced the control over who can access their websites to a company, with no second thoughts whatsoever.
I worked at a fully remote company that did the best job hiring juniors in my 20 year career. The talent and enthusiasm in that pool was great and really injected something into teams.
What changed was ZIRP ending. The layoffs from that were real, and the managers who can't hire a ton angle for more senior people instead. The junior culture changed overnight.
Is it me or are people too eager to "one track mind" everything into AI? If I had said thirty years ago that Microsoft would remote disable old copies of Office asking you to upgrade, literally no one would be surprised. This is standard MO for Microsoft, even in a world without AI.
The main business is selling the data. You use Speedtest.net to troubleshoot your connection but metrics captured with the test alongside location data give telcos invaluable insights on where they should improve their networks. Telcos pay 6 figures annually for this data and we have a few hundreds of of those big MNOs globally. This market is pretty big. Accenture is in trouble with their main consulting business due to AI so acquiring data business is one of the smart strategies they can implement to stay relevant.
To all commenters who think they can code it over the weekend, yes you are right. I coded my first speed checker over the weekend in 2008 but it took me 18 years to grow the user base , figure out entreprise sales strategy and exit. Its not easy as it seems.