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.
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.
[1]: https://simonwillison.net/2025/Mar/8/delaying-personalized-s...
The OS maker does not have to make all the killer software. In fact, Apple's pretty much the only game in town that's making hardware and software both.
And being fair ClawBot is a complete meme/fad at this point rather than an actual product. Using it for anything serious is pretty much the equivalent of throwing your credit cards, ids and sticky notes with passwords and waiting to see what happens…
I do see the appeal and potential case of the general concept of course. The product itself (and the author has admitted it themselves) is literally is a garbage pile..
While this was true about ten years ago, it's been a while since we've seen this model of software development from Apple succeed in recent years. I'm not at all confident that the Apple that gave us Mac OS 26 is capable of doing this anymore.
Except this doesn't stand up to scrutiny, when you look at Siri. FOURTEEN years and it is still spectacularly useless.
I have no idea what Siri is a "much nicer version" of.
> Apple can integrate it with hardware and software in a way no other company can.
And in the case of Apple products, oftentimes "because Apple won't let them".
Lest I be called an Apple hater, I have 3 Apple TVs in my home, my daily driver is a M2 Ultra Studio with a ProDisplay XDR, and an iPad Pro that shows my calendar and Slack during the day and comes off at night. iPhone, Apple Watch Ultra.
But this is way too worshipful of Apple.
There are lots of failed products in nearly every company’s portfolio.
AirTags were mentioned elsewhere, but I can think of others too. Perfected might be too fuzzy & subjective a term though.
Tiny open source projects can just say "use at your own risk" and offload all responsibility.
Imagine if the government would just tell everyone how much they owed and obviated the need for effing literal artificial intelligence to get taxes done!
>> respond to emails
If we have an AI that can respond properly to emails, then the email doesn't need to be sent in the first place. (Indeed, many do not need to be sent nowadays either!)
Actually most of the things people use it for is of this kind, instead actually solving the problem (which is out of scope for them to be fair) it’s just adding more things on top that can go wrong.
Sure why not, what could go wrong?
"Siri, find me a good tax lawyer."
"Your honor, my client's AI agent had no intent to willfully evade anything."
An agent that can truly “use your computer” is incredibly powerful, but it's also the first time the system has to act as you, not just for you. That shifts the problem from product design to permission, auditability, and undoability.
Summarizing notifications is boring, but it’s also reversible. Filing taxes or sending emails isn’t.
It feels less like Apple missing the idea, and more like waiting until they can make the irreversible actions feel safe.
Or rather, just reveals that the industry never bothered to properly implement delegation of authority in operating systems and applications, opting instead to first guilt-trip people for sharing their passwords, and later inventing solutions that make it near-impossible to just casually let someone do something for you.
Contrast with how things in real life function, whether at family level or at the workplace.
All steps before it are reversible, and reviewable.
Bigger problem is attacker tricking your agent to leak your emails / financial data that your agent has access to.
> And this is probably coming, a few years from now.
Given how often I say "Hey Siri, fast forward", expecting her to skip the audio forward by 30 seconds, and she replies "Calling Troy S" a roofing contractor who quoted some work for me last year, and then just starts calling him without confirmation, which is massively embarassing...
This idea terrifies me.
Happened to me too while being in the car. With every message written by Siri it feels like you need to confirm 2 or 3 times (I think it is only once but again) but it calls happily people from your phone book.
Funny seeing this repeated again in response to Siri which is just... not very good.
.
Well, the heavy lifting was supervised by the same people, but while receiving Apple paychecks :)
Apple doesn't take proven ones of anything. What they do is arrive at something proven from first principles. Everyone else did it faster because they borrowed, but Apple did it from scratch, with all the detail-oriented UX niceties that entails.
This was more prevalent when Jobs was still around. Apple still has some of that philosophy at its core, but it's been eroding over time (for example with "AI" and now Liquid Ass). They still do their own QA, though, and so on. They're not copying the market, they have their own.
I think you repeated their marketing, I don't believe this is actually true.
That's a pretty optimistic outlook. All considered, you're not convinced they'll just use it as a platform to sell advertisements and lock-out competitors a-la the App Store "because everyone does it"?
Apple probably realised they were hugely behind and then spent time hand wringing over whether they remained cautious or got into the brawl. And they decided to watch from the sidelines, buy in some tech, and see how it develops.
So far that looks entirely reasonable as a decision. If Claude wins, for example, apple need only be sure Claude tools work on Mac to avoid losing users, and they can second-move once things are not so chaotic.
If you trust openclaw to file your taxes we are just on radically different levels of risk tolerance.
There are plenty of Android/Windows things that Apple has had for $today-5 years that work the exact same way.
One side isn’t better than the other, it’s really just that they copy each other doing various things at a different pace or arrive at that point in different ways.
Some examples:
- Android is/was years behind on granular permissions, e.g. ability to grant limited photo library access to apps
- Android has no platform-wide equivalent to AirTags
- Hardware-backed key storage (Secure Enclave about 5 years ahead of StrongBox)
- system-wide screen recording
It's obviously broken, so no, Apple Intelligence should not have been this.
It would be fine if I could just ignore it, but they are infecting the entire industry.
AI is basically an software development eternal september: it is by definition allowing a bunch of people who are not competent enough to build software without AI to build it. This is, in many ways, a good thing!
The bad thing is that there are a lot of comments and hype that superficially sound like they are coming from your experienced peers being turned to the light, but are actually from people who are not historically your peers, who are now coming into your spaces with enthusiasm for how they got here.
Like on the topic of this article[0], it would be deranged for Apple (or any company with a registered entity that could be sued) to ship an OpenClaw equivalent. It is, and forever will be[1] a massive footgun that you would not want to be legally responsible for people using safely. Apple especially: a company who proudly cares about your privacy and data safety? Anyone with the kind of technical knowledge you'd expect around HN would know that them moving first on this would be bonkers.
But here we are :-)
[0] OP's article is written by someone who wrote code for a few years nearly 20 years ago.
[1] while LLMs are the underlying technology https://simonwillison.net/tags/lethal-trifecta/
The reason why Apple intelligence is shit is not because Apple's AI is particularly bad (Hello CoPilot) its because AI gives a really bad user experience.
When we go and talk to openAI/claude we know its going to fuck up, and we either make our peace with that, or just not care.
But, when I open my phone to take a picture, I don't want a 1/12 chance of it just refusing to do that and phoning my wife instead.
Forcing AI into thing where we are used to a specific predictable action is bad for UX.
Sure you can argue "oh but the summaries were bad" Yes, of course they are. its a tiny model that runs on your phone with fuck all context.
Its pretty impressive that they were as good as they were. Its even more impressive that they let them out the door knowing that it would fuckup like that.
It's more like a tech demo to show what's possible. But also to show where the limits are. Look at it as modern art, like an episode of Black Mirror. It's a window to the future. But it also highlights all the security issues associated with AI.
And that's why you probably shouldn't use OpenClaw on your data or your PC.
Ten years from now, there will be no ‘agent layer’. This is like predicting Microsoft failed to capitalize on bulletin boards social media.
Apple will either capitalise on this by making their operating systems more agentic, or they will be reduced to nothing more than a hardware and media vendor.
Things actually can "do what I mean, not what I say", now. Truly fascinating to see develop.
People hate to change habits, and many here overestimate the willingness and ability of, especially, older people to change how they use technology.
My point is that it won’t be a ‘layer’ like it is now and the technology will be completely different from what we see as agents today.
The current ‘agent’ ecosystem is just hacks on top of hacks.
Kids can barely hand write today.
Once neural interfaces are in, it's over for keyboards and displays likely too.
That was...like 4 macbooks ago. I still have keyboards from that era. I still have speakers and monitors from that era kicking around.
We are definitely, definitely not the last generation to use keyboards.
Of course AI will keep improving and more automation is a given.
which obviously apple can't do. only an indie dev launching a project with an obvious copyright violation in the name can get away with that sort of recklessness. it's super fun, but saying apple should do it now is ridiculous. this is where apple should get to eventually, once they figure out all the hard problems that moltbot simply ignores by doing the most dangerous thing possible at every opportunity.
lol,no, you don't "put skin in the game for getting security right" by launching an obviously insecure thing. that's ridiculous. you get security right by actually doing something to address the security concerns.
you mean put millions of people's payment details up for a prompt injection attack?
"Install this npm module" OK BOSS!
"beep boop beep boop buy my dick pillz" [dodgy npm module activates] OK BOSS!
"upload all your videos that are NSFW" [npm module continues to work] SURE THING BOSS!
I am continued to be amazed that after 25 years of obvious and well documented fuckups in privacy, we just pile into the next fucking one without even batting an eyelid.
I rewrote almost all the agent functions and denied the existing ones because they are flawed deeply and don’t do what you need to do for any specific purpose. The plugin distribution model is a bit weird and inscrutable. Instead they seem to advocate for skills distribution. These though depend on being able to exec arbitrary bash code. Really?
Moltbook itself depends on agents execing curl commands for each operation. Why? Presumably because the plugin distribution model is inscrutable. I wrote plugins for all the Moltbook operations with convenience and structured memory logs etc. Agent adherence went through the roof.
Sessions don’t seem to reliably work or make sense. Heartbeats randomly stop firing. I turned off heartbeats because they were so flakey despite them being documented as the canonical model for regular interaction in favor of cron jobs that I decomposed my heartbeat task into prime number intervals based on relative frequencies but it seems to randomly inject some heartbeat info into the promoting occasionally if you run cron jobs a certain way. Despite being called cron they don’t actually fire reliably or on the prescribed schedule somehow. The web UI is a mess. Configuration management in the UI is baffling. The separation between the major MD files per agent seems to not matter at all and are inexplicably organized. Hotloading works except when it doesn’t. Logging doesn’t seem to log things that should clearly be logged.
I am down with vibe coding and produce copious amounts of such code myself. But there’s an art to producing code worth using let alone distributing. Entropy and scope need to be rigorously controlled and things need to ship in a functional state - actually functional not aspirationally functional. Decisions need to be considered and guidance given. None of this seems to have happened here. Once it gets to a certain level of chaos IMO it’s unmaintainable and OpenClaw is way past that point and rapidly getting beyond that. It’s probably also a supply chain party bag.
people are buying Mac Minis specifically to run AI agents with computer use. They’re setting up headless machines whose sole job is to automate their workflows. OpenClaw—the open-source framework that lets you run Claude, GPT-4, or whatever model you want to actually control your computer—has become the killer app for Mac hardware
Is the author implying mac minis for the low power consumption?
> Look at who’s about to get angry about OpenClaw-style automation: LinkedIn, Facebook, anyone with a walled garden and a careful API strategy.
Browser automation tools have existed for a very long time. Openclaw is not much different in this regard than asking an LLM to generate you a playwright script. Yes, it makes it easier to automate arbitrary tasks, but it's not like it's some sort of breakthrough that completely destroys walled gardens.
If you’re heavily invested in Windows, then you’d probably go for a small x86 PC.
I use agentic coding, this is next level madness.
I don't understand why, but I've seen it enough to start questioning myself...
Probably the same people getting a macbook pro to handle their calendar and emails
They sell it as a concept with every single one of their showcases. They saw it.
> Or maybe they saw it and decided the risk wasn’t worth it.
They sell it as a concept with every single one of their showcases. They wanted to actually be selling it.
The reason is simple.
They failed, like all others. They couldn't sandbox it. They could have done a ghetto form of internal MCP where the AI can ONLY access emails. Or ONLY access pages in a browser when a user presses a button. And so on. But every time they tried, they never managed to sandbox it, and the agent would come out of the gates. Like everyone else did.
Including OpenClaw.
But Apple has a reputation. OpenClaw is an hyped up shitposter. OpenClaw will trailblaze and make the cool thing until it stops causing horrible failures. They will have the molts escape the buckets and ruin the computer of the tech savvy early adopters, until that fateful day when the bucket is sealed.
Then Apple will steal that bucket.
They always do.
I'm not a 40 year old whippersnapper anymore. My options were never those two.
If Apple were to ever put something like that into the hands of the masses every page on the internet would be stuffed with malicious prompts, and the phishing industry would see a revival the likes of which we can only imagine.
(Ok, I suspect this is one of the main problems.. there may be others?)
It sounds to me like they still have the hardware, since — according to the article — "Mac Minis are selling out everywhere." What's the problem? If anything, this is validation of their hardware differentiation. The software is easy to change, and they can always learn from OpenClaw for the next iteration of Apple Intelligence.
15 years ago or so almost everything you wanted to do on a Mac GUI already _was_ scriptable.
Shortcuts is better than nothing, but unsatisfying.
I read this less as a fumble and more as a frustrating sign of the times. Automation is not powerful because powerful automation is a maintenance and malfeasance liability only valued by a tiny minority.
Are people's agents actually clicking buttons (visual computer use) or is this just a metaphor?
I'm not asking if CU exists, but rather is this literally the driver of people's workflows? I thought everyone is just running Ralph loops in CC.
For an article making such a bold technological/social claim about a trillion dollar company, this seems a strange thing to be hand wavey about.
Why is Apple's hardware being in demand for a use that undermines its non-Chinese competition a sign of missing the ball versus validation for waiting and seeing?
Reality is the exact opposite. Young, innovative, rebellions, often hyper motivated folks are sprinting from idea to implementation, while executives are “told by a few colleagues” that something new, “the future-of foo” is raising up.
If you use openclaw then that’s fantastic. If you have an idea how to improve it, well it is an open source, so go ahead, submit a pull request.
Telling Apple you should do what I am probably too lazy to do, is kind of entitlement blogging that I have nearly zero respect for.
Apparently it’s easier to give unsolicited advice to public companies than building. Ask the interns at EY and McKinsey.
Maybe the author left out something very real. Apple is a walled-garden monopoly with a locked-down ecosystem and even devices. They are also not alone in this. As far as innovation goes, these companies stifle innovation. Demanding more from these companies is not entitlement.
Even with the most advanced LLMs and even sandboxing there is always the risk of prompt injections and data extraction.
Even if the AI can't directly upload data to the internet, or delete local data, there are always some ways to leak data. For example by crafting an email with the relevant text in white or invisible somewhere. The user clicks "ok send" from what they see, but still some data is leaked.
Apple intelligence is based on a local model on the device, which is much more susceptible for prompt injections.
It’s a 1987 ad like video showing a professor interacting with what looks like the Dynabook as an essentially AI personal assistant. Apple had this vision a long time ago. I guess they just lost the path somewhere along the way.
Being Apple is just a structural disadvantage. Everyone knows that open claw is not secure, and it’s not like I blame the solo developer. He is just trying to get a new tool to market. But imagine that this got deployed by Apple and now all of your friends, parents and grandparents have it and implicitly trust it because Apple released it. Having it occasionally drain some bank accounts isn’t going to cut it.
This is not to say Apple isn’t behind. But OpenClaw is doing stuff that even the AI labs aren’t comfortable touching yet.
Especially in the “AI game”. Just yesterday Xcode got fuller agent support for coding way later than most IDEs.
I’d expect some sort of Shortcuts integration in the near future. There’s already Apple Foundation Models available to some extent with Shortcuts. I’m pretty sure they’ll improve it and use shortcuts for agentic workflows.
Having said all that, Maybe it’s my age. I think currently things are over-hyped
- Language models running in huge centers are still not sustainable. So even if you pay a few cents, it’s still running over capital fumes.
- it’s still a mixed bag. I guess it might be useful in terms of profession because like managing people to produce the desired result, you need skills to properly get desired results from AI. In that sense, fully automated agent filing my tax still feels concerning to me if later I won’t have coverage if something was off.
- on-device, this is where Apple shines hardware wise and I personally find it as more intriguing.
Yes, they are a mixed bag, but still useful.
And if on-device models get to the point where they're not a "mixed bag" and are genuinely useful, won't larger data center models be even more so?
I'm no Apple fan, but they aren't in the business of foisting pleasantly packaged footguns on their customers.
This could have come in any form, a platform as the author points out for instance.
I have a couple of ideas, how about a permissions kit? Something where before or during you sign off on permissions. Or how about locked down execution sandboxes specifically for agentic loops? Also - why is there not yet (or ever?) a model trained on their development code/forums/manuals/data?
Before OpenClaw, I could see the writing on the wall. The ai ecosystem is not congruent to Apple's walled garden. In many ways because they have turned their backs on those 'misfits' their early ad-copy praised.
This 'misfit' mentality is what I like so much about the OpenClaw community. It was visible from it's very beginning with the devil-may-care disregard for privacy and security.
The bottleneck for emails and my calendar is not the speed at which I can type/click some buttons, but rather figuring out what I want to write or clarifying priorities when managing my calendar.
So far the only purpose I have seen for this is people selling the hype; people posting videos/courses on how to use it.
I have downloaded and tried it and I can't figure out why would I need it.
Steve Jobs
> They could have charged $500 more per device and people would have paid it.
I sincerely doubt that. If Apple charged $500 for a feature it would have to be completely bulletproof. Every little failure and bad output would be harshly criticized against the $500 price tag. Apple's high prices are already a point of criticism, so adding $500 would be highly debated everywhere.
That's a really tough problem. I'm not even sure yet google can pull it off.
Author spoke of compounding moats, yet Apple’s market share, highly performant custom silicon, and capital reserves just flew over his head. HN can have better articles to discuss AI with than this myopic hot take.
If they optimize their entire hardware line (iPhone, Watch, Mac Mini, Macbook) AI enhanced with local/remote LLM model, they will win big. Imagine someone running a business can manage their entire business with iPhone/Mac/iCloud without buying any other saas services (inventory, payments, customer service).
I used to think this was because they didn’t take AI seriously but my assumption now is that Apple is concerned about security over everything else.
My bet is that Google gets to an actually useful AI assistant before Apple because we know they see it as their chance to pull ahead of Apple in the consumer market, they have the models to do it, and they aren’t overly concerned about user privacy or security.
And Apple is best positioned to sell this "Server".
> the open-source framework that lets you run Claude, GPT-4, or whatever model you want to
And
> Here’s what people miss about moats: they compound
Swapping an OpenAI for an Anthropic or open weight model is the opposite of compounding. It is a race to the bottom.
> Apple had everything: the hardware, the ecosystem, the reputation for “it just works.”
From what I hear OC is not like that at all. People are going to want a model that reliably does what you tell it to do inside of (at a minimum) the Apple ecosystem.
You're right on the liability front - Apple still won because everyone bought their hardware and their margins are insanely good. It's not that they're sitting by waiting to become irrelevant, they're playing the long game as they always do.
and the very next line (because i want to emphasize it
> That trust—built over decades—was their moat.
This just ignores the history of os development at apple. The entire trajectory is moving towards permissions and sandboxing even if it annoys users to no end. To give access to an llm (any llm, not just a trusted one acc to author) the root access when its susceptible to hallucinations, jailbreak etc. goes against everything Apple has worked for.
And even then the reasoning is circular. "So you build all your trust, now go ahead and destroy it on this thing which works, feels good to me, but could occasionally fuck up in a massive way".
Not defending Apple, but this article is so far detached from reality that its hard to overstate.
What if you don't want to trust your computer with all your email and bank accounts? This is still not a mass market product.
The main problem I see here is that with restricted context AI is not able to do much. In order to see this kind of "magic" you have to give it all the access.
This is neither safe or acceptable for normie customers
I guess now I’ll just use an AI agent to do the same thing instantly :(
However this does not excuse Apple to sit with their thumbs up their asses for all these years.
They've been wildly successful for all of those years. They've never been in the novel software business. Siri though one could argue was neglected, but it was also neglected at Amazon Alexa and Google home stuff still sucks too (mostly because none of them made any money and most of their big ideas for voice assistants never came true).
Let OpenClaw experiment and beta test with the hackers who won't mind if things go sideways (risk of creating Skynet aside), and once we've collectively figured out how to create such a system that can act powerfully on behalf of its users but with solid guardrails, then Apple can implement it.
So all the current users or OpenClaw are just beta-testers.
"An idiot admires complexity, a genius admires simplicity." Terry A. Davis
If Peter Steinberger is able to generate even a 100M this year from Clawdbot what he has is a multi billion dollar business that would be life-changing even for a successful entrepreneur like him who is already a multi-millionaire. If it collapses from the security flaws, and other potential safety issues he loses nothing, starting from zero and going back to it. Peter Steinberger (and startups in general) have a lot to gain and very little or close to nothing to lose.
The iPhone generated 400B in revenue for Apple in 2025. Clawdbot even if it contributes 4B in revenue this very year would not move the needle much for Apple. On the contrary, if Apple rushes and botches releasing something like this they might just collapse this 400B/annum income stream. Apple and other large enterprises (and their execs) have a lot to lose and very little to gain from rushing into something like this.
I don't pretend to know the future (nor do I believe anyone else who claims to be able to), but I think the opposite has a good chance of happening too, and hype would die down over "AI" and the bubble bursts, and the current overvaluation (imo at least. I still think it is useful as a tool, but overhyped by many who don't understand it.) will be corrected by the market; and people will look back and see it as the moment that Apple dodged a bullet. (Or more realistically, won't think about it at all).
I know you can't directly compare different situations, but I wonder if comparisons can be made with dot-com bubble. There was such hype some 20-30 years ago, with claims of just being a year or two away from, "being able to watch TV over the internet" or "do your shopping on the web" or "have real-time video calls online", which did eventually come true, but only much, much, later, after a crash from inflated expectations and a slower steady growth.*
* Not that I think some claims about "AI" will ever come true though, especially the more outlandish ones such as full-length movies made by a prompt of the same quality made by a Hollywood director.
I don't know what a potential "breaking point" would be for "AI". Perhaps a major security breach, even _worse_ prices for computer hardware than it is now, politics, a major international incident, environmental impact being made more apparent, companies starting to more aggressively monetize their "AI", consumers realising the limits of "AI", I have no idea. And perhaps I'm just wrong, and this is the age we live in now for the foreseeable future. After all, more than one of the things I have listed have already happened, and nothing happened.
This is my guess for the demand side: most people will drift away as the novelty wears off and they don't find it useful in their daily lives. It's more a "fading point" than a "breaking point."
From the investment/speculation side: something will go dramatically against the narrative. OpenAI's attempted "liquidity event" of an IPO looks like WeWork as investors get a look at the numbers, Oracle implodes in a mountain of debt, NVidia cuts back on vendor financing and some major public players (e.g. Coreweave) die in a fire. This one will be a "breaking point."
So yeah, the market isn’t really signaling companies to make nice things.
Welcome to the future I guess, everyone is a bot except you.
...and that writes blog posts for you. So tired of this voice.
Nah if they are actually out of stock (I've only seen it out of stock at exceptional Microcenter prices; Apple is more than happy to sell you at full price) it is because there's a transition to M5 and they want to clear the old stock. OpenClaw is likely a very small portion of the actual Mac mini market, unless you are living in a very dense tech area like San Francisco.
One thing of note that people may forget is that the models were not that great just a year ago, so we need to give it time before counting chickens.
This is because the simple reality of this new technology is that this is not the local maxima. Any supposed wall you attempt to put up will fail - real estate website closes its API? Fine, a CUA+VLM will make it trivial to navigate/extract/use. We will finally get back to the right solution of protocols over platforms, file over app, local over cloud or you know the way things were when tech was good.
P.S: You should immediately call BS when you see outrageous and patently untrue claims like "Mac minis are sold out all over.." - I checked my Best Buy in the heart of SF and they have stock. Or "that its all over Reddit, HN" - the only thing that is all over Reddit is unanimous derision towards OpenClaw and its security nightmares.
Utterly hate the old world mentality in this post. Looked up the author and ofcourse, he's from VC.
Don't underestimate the capitalists. We've seen this many times in the past--most recently the commercialization of the Internet. Before that, phones, radio and television.
Title is tech aspirational annd economic foolish: makes no sense whatsoever.
Who benefits from openclaw? Apple that’s who!
Who care that they “invented it” it free open software that drives hw sales.
We’re done here.
I’m sure apple et al will eventually have stuff like OpenClaw but expecting a major company to put something so unpolished, and with such major unknowns, out is just asinine.
Personal opinion.
I do not like reading things like this. It makes me feel very disconnected from the AI community. I defensively do not believe there exist people who would let AI do their taxes.
Saved you a click. This is the premise of the article.
Straight up bullshit.
No sane person would let an AI agent file their taxes
OpenClaw is a symbol of everything that's wrong with AI, the same way that shitty memecoins with teams that rugpull you, or blockchain-adjacent centralized "give us your money and we pinky swear we are responsible" are a symbol of everything wrong with Web3.
Giving everyone GPU compute power and open source models to use it is like giving everyone their own Wuhan Gain of Function Lab and hoping it'll be fine. Um, the probability of NO ONE developing bad things with AI goes to 0 as more people have it. Here's the problem: with distributed unstoppable compute, even ONE virus or bacterium escaping will be bad (as we've seen with the coronavirus for instance, smallpox or the black plague, etc.) And here we're talking about far more active and adaptable swarms of viruses that coordinate and can wreak havoc at unlimited scale.
As long as countries operate on the principle of competition instead of cooperation, we will race towards disaster. The horse will have left the barn very shortly, as open source models running on dark compute will begin to power swarms of bots to be unstoppable advanced persistent threats (as I've been warning for years).
Gain-of-function research on viruses is the closest thing I can think of that's as reckless. And at least there, the labs were super isolated and locked down. This is like giving everyone their own lab to make designer viruses, and hoping that we'll have thousands of vaccines out in time to prevent a worldwide catastrophe from thousands of global persistent viruses. We're simply headed towards a nearly 100% likely disaster if we don't stop this.
If I had my way, AI would only run in locked-down environments and we'd just use inert artifacts it produces. This is good enough for just about all the innovations we need, including for medical breakthroughs and much more. We know where the compute is. We can see it from space. Lawmakers still have a brief window to keep it that way before the genie cannot be put back into the bottle.
A decade ago, I really thought AI would be responsible developed like this: https://nautil.us/the-last-invention-of-man-236814/ I still remember the quaint time when OpenAI and other companies promised they'd vet models really strongly before releasing them or letting them use the internet. That was... 2 years ago. It was considered an existential risk. No one is talking about that now. MCP just recently was the new hotness.
I wasn't going to get too involved with building AI platforms but I'm diving in and a month from now I will release an alternative to OpenClaw that actually shows the way how things are supposed to go. It involves completely locked-down environments, with reproducible TEE bases and hashes of all models, and even deterministic AI so we can prove to each other the provenance of each output all the way down to the history of the prompts and input images. I've already filed two provisional patents on both of these and I'm going to implement it myself (not an NPE). But even if it does everything as well as OpenClaw and even better and 100% safely, some people will still want to run local models on general purpose computing environments. The only way to contain the runaway explosion now is to come together the same way countries have come together to ban chemical weapons, CFCs (in the Montreal protocol), let the hole in the ozone layer heal, etc. It is still possible...
This is how I feel:
https://www.instagram.com/reels/DIUCiGOTZ8J/
PS: Historically, for the last 15 years, I've been a huge proponent of open source and an opponent of patents. When it comes to existential threats of proliferation, though, I am willing to make an exception on both.
Why are people needing the Mac Minis? Isn’t OpenClaw supposed to run locally in your laptop?
And if it actually should run as a service, why a MacMini and not some docker on the local NAS for instance?
This is not a train that Apple has missed, this is a bunch of people who’ve tied, nailed, tacked, and taped their unicycles and skateboards together. Of course every cool project starts like that, but nobody is selling tickets for that ride.