It's a tale as old as time that developers, particularly junior developers, are convinced they could "slap together something in one weekend" that would replace expensive SAAS software and "just do the parts of it we actually use". Unfortunately, the same arguments against those devs regular-coding a bespoke replacement apply to them vibe-coding a bespoke replacement: management simply doesn't want to be responsible for it. I didn't understand it before I was in management either, but now that I'm in management I 100% get it.
"Thankfully, we live in a beautifully democratic and capitalistic society where we can fight in court."
Of course he's "thankful" for that, since in our "beautifully democratic and capitalistic" society, Flock can use their $658 million of VC funding [1] to wage lawfare against the have-nots with their armies of lobbyists and lawyers. [2]
The biggest issue I see is Microsoft's entire mentality around AI adoption that focuses more on "getting the numbers up" then actually delivering a product people want to use.
Most of the announcements I hear about Copilot, it's always how they've integrated it into some other piece of software or cut a deal with yet another vendor to add it to that vendors product offering. On the surface there's nothing wrong with doing that but that just seems to be the ONLY thing Microsoft is focused on.
Worse yet, most of these integrations seem like a exercise in ticking boxes rather than actually thinking through how integrating Copilot into a product will actually improve user experience. A great example was someone mentioned that Copilot was now integrated into the terminal app but beyond an icon + a chat window, there is zero integration.
Overall, MS just reeks of an organization that is cares more about numbers on a dashboard and pretty reports than they are on what users are actually experiencing.
On Firefox, web accessible resources are available at "moz-extension://<extension-UUID>/myfile.png" <extension-UUID> is not your extension's ID. This ID is randomly generated for every browser instance. This prevents websites from fingerprinting a browser by examining the extensions it has installed. https://developer.mozilla.org/en-US/docs/Mozilla/Add-ons/Web...
> This is exactly what Apple Intelligence should have been... They could have shipped an agentic AI that actually automated your computer instead of summarizing your notifications. Imagine if Siri could genuinely file your taxes, respond to emails, or manage your calendar by actually using your apps, not through some brittle API layer that breaks every update.
And this is probably coming, a few years from now. Because remember, Apple doesn't usually invent new products. It takes proven ones and then makes its own much nicer version.
Let other companies figure out the model. Let the industry figure out how to make it secure. Then Apple can integrate it with hardware and software in a way no other company can.
Right now we are still in very, very, very early days.
This is an industry we're[0] in. Owning is at one end of the spectrum, with cloud at the other, and a broadly couple of options in-between:
1 - Cloud – This is minimising cap-ex, hiring, and risk, while largely maximising operational costs (its expensive) and cost variability (usage based).
2 - Managed Private Cloud - What we do. Still minimal-to-no cap-ex, hiring, risk, and medium-sized operational cost (around 50% cheaper than AWS et al). We rent or colocate bare metal, manage it for you, handle software deployments, deploy only open-source, etc. Only really makes sense above €$5k/month spend.
3 - Rented Bare Metal – Let someone else handle the hardware financing for you. Still minimal cap-ex, but with greater hiring/skilling and risk. Around 90% cheaper than AWS et al (plus time).
4 - Buy and colocate the hardware yourself – Certainly the cheapest option if you have the skills, scale, cap-ex, and if you plan to run the servers for at least 3-5 years.
A good provider for option 3 is someone like Hetzner. Their internal ROI on server hardware seems to be around the 3 year mark. After which I assume it is either still running with a client, or goes into their server auction system.
Options 3 & 4 generally become more appealing either at scale, or when infrastructure is part of the core business. Option 1 is great for startups who want to spend very little initially, but then grow very quickly. Option 2 is pretty good for SMEs with baseline load, regular-sized business growth, and maybe an overworked DevOps team!
I don’t believe this was ever confirmed by Apple, but there was widespread speculation at the time[1] that the delay was due to the very prompt injection attacks OpenClaw users are now discovering. It would be genuinely catastrophic to ship an insecure system with this kind of data access, even with an ‘unsafe mode’.
These kinds of risks can only be _consented to_ by technical people who correctly understand them, let alone borne by them, but if this shipped there would be thousands of Facebook videos explaining to the elderly how to disable the safety features and open themselves up to identity theft.
The article also confuses me because Apple _are_ shipping this, it’s pretty much exactly the demo they gave at WWDC24, it’s just delayed while they iron this out (if that is at all possible). By all accounts it might ship as early as next week in the iOS 26.4 beta.
I'd actually say the opposite is the case. B2B (even SaaS) is probably the most robust when it comes to AI resistance. The described "in house vibe coded SaaS replacement" does not mirror my experience in B2B at all. The B2B software mindset I've encountered the most is "We'll pay you so we don't have to wrestle with this and can focus on what we do. We'll pay you even more if we worry even less." which is basically the opposite of...let's have someone inhouse vibe code and push to production. B2B is usually fairly conservative.
All these tools to build something, but nothing to build. I feel like I am part of a Pyramid Scheme where every product is about building something else, but nothing reaches the end user.
Note: nothing against fluid.sh, I am struggling to figure out something to build.
Just tested the new Opus 4.6 (1M context) on a fun needle-in-a-haystack challenge: finding every spell in all Harry Potter books.
All 7 books come to ~1.75M tokens, so they don't quite fit yet. (At this rate of progress, mid-April should do it ) For now you can fit the first 4 books (~733K tokens).
Results: Opus 4.6 found 49 out of 50 officially documented spells across those 4 books. The only miss was "Slugulus Eructo" (a vomiting spell).
We don't vary our model quality with time of day or load (beyond negligible non-determinism). It's the same weights all day long with no quantization or other gimmicks. They can get slower under heavy load, though.
I spent a good part of my career (nearly a decade) at Google working on getting Clang to build the linux kernel. https://clangbuiltlinux.github.io/
This LLM did it in (checks notes):
> Over nearly 2,000 Claude Code sessions and $20,000 in API costs
It may build, but does it boot (was also a significant and distinct next milestone)? (Also, will it blend?). Looks like yes!
> The 100,000-line compiler can build a bootable Linux 6.9 on x86, ARM, and RISC-V.
The next milestone is:
Is the generated code correct? The jury is still out on that one for production compilers. And then you have performance of generated code.
> The generated code is not very efficient. Even with all optimizations enabled, it outputs less efficient code than GCC with all optimizations disabled.
I'm now in my 50s. I tried management but prefer working as an IC. I think I'm good but I know most companies would never hire me. One thing I do now is try to look after all the youngest grads and new joiners. Its so cutthroat now it seems no one has time to help anyone else, so I like helping people get up and running and encouraging them to enjoy their work while being productive and getting their skills up. No one else seems to care.
There aren't any "AI" products that have enough value.
Compare to their Office suite, which had 100 - 150 engineers working on it, every business paid big $$ for every employee using it, and once they shipped install media their ongoing costs were the employees. With a 1,000,000:1 ratio of users to developers and an operating expense (OpEx) of engineers/offices/management. That works as a business.
But with "AI", not only is it not a product in itself, it's a feature to a product, but it has OpEx and CapEx costs that dominate the balance sheet based on their public disclosures. Worse, as a feature, it demonstrably harms business with its hallucinations.
In a normal world, at this point companies would say, "hmm, well we thought it could be amazing but it just doesn't work as a product or a feature of a product because we can't sell it for enough money to both cover its operation, and its development, and the capital expenditures we need to make every time someone signs up. So a normal C staff would make some post about "too early" or whatever and shelve it. But we don't live in a normal world, so companies are literally burning the cash they need to survive the future in a vain hope that somehow, somewhere, a real product will emerge.
Hopefully this is a wakeup call to the software engineers and other employees at those companies - it's no longer a hypothetical future where the tools you are building might be abused, it's today.
OTOH, I was hired by an enterprise that was many months into a giant backend rewrite. After wrapping my head around the many plans, I realized they were rewriting Django, badly. One weekend I prototyped the whole thing… in Django. It worked. It met the specs. It was a CRUD app with a REST API.
I came in to work Monday morning, showed it off, and inadvertently triggered a firestorm. Later my boss told me not to do that again because it caused havoc with schedules and such.
So I quit and found a better job. Sometimes the new guy can make a better version themselves over the weekend, not because they’re a supergenius, but because they’re not hampered by 47 teams all trying to get their stamp on the project.
(In before “prime example of overconfidence!”: feel free to doubt. It was a CRUD app with a handful of models on a PostgreSQL backend. They were writing a new Python web framework to serve it, complete with their own ORM and forms library and validation library. Not because the existing ones wouldn’t work, mind you, but more out of not realizing that all these problems were already sufficiently solved for their requirements.)
This is really pushing it, considering it’s trained on… internet, with all available c compilers. The work is already impressive enough, no need for such misleading statements.
> Around 90% of superstar adults had not been superstars as children, while only 10% of top-level kids had gone on to become exceptional adults (see chart 1). It is not just that exceptional performance in childhood did not predict exceptional performance as an adult. The two were actually negatively correlated, says Dr Güllich.
Even if "only" 10% of elite kids go on to become elite adults, 10% is orders of magnitude larger than the base percentage of adults who are elite athletes, musicians, etc. This doesn't sound "uncorrelated" to me so much as "not as strongly correlated as one might expect."
And describing something that happens 10% of the time as "rare" sounds a bit weird, like referring to left-handedness (also about 1 in 10) as rare.
Amateur. Opus 4.6 this afternoon built me a startup that identifies developers who aren’t embracing AI fully, liquifies them and sells the produce for $5/gallon. Software Engineering is over!
"...and then unfortunately there is terroristic organizations like DeFlock, whose primary motivation is chaos. They are closer to Antifa than they are anything else."
"We're not forcing Flock on anyone..."
It is a short 1:32 video, I encourage people to watch it for themselves.
I thought DeFlock was just publishing locations of cameras and lawfully convincing local governments to not use Flock, primarily through FOIA requests.
Many years ago, I worked at a company with a product that ran on Mac and Windows. The Mac version was pretty solid, but the Windows version had some problems.
They had a talented team of developers who were mostly Mac experts and just starting to get a grip on Windows.
I was known at the time as a "Windows expert", so they hired me to help the team get the Windows version into shape.
My typical day started with "house calls". People would ping me with their Windows questions and I'd go door to door to help solve them - and to make sure they understood how to do things on Windows.
In the afternoon, I would work on my own code, but I told everyone they could always call on me for help with a Windows problem, any time of day.
One colleague asked me: "Mike, how can you afford to be so generous with your time?"
Then in a performance review, I got this feedback:
"Mike, we're worried. Your productivity has been OK lately, but not great. And it's surprising, because the productivity of the rest of the team has improved a lot during this time."
I bit my tongue, but in retrospect I should have said:
