When this first landed I asked what the fix could even be. Everyone needs a new ID at a minimum. But then I got to thinking: 1) is that the point? Conspiratorial thinking I know but “hey all Our ids got hacked I guess we need a national id”. And related 2) the current id system from a security standpoint was a band aid fix for outdated world to be shoehorned into a modern one. IDscan was never cryptographic proof you were who you said you were. Maybe better than “enter your name and SSN” but bottom line, at least in US there is no cryptographically secure identity system that proves you are the citizen you say. And that fact bleeds into all sorts of patchwork solutions, fraud, etc. Moreover there are serious philosophical hurdles to getting to one. I’m not even positive I want one. But unless there is some zero-trust way to do this, I’m not sure what the fix would be.

And then, in real life, one discovers that institutions route around in creative ways for all sorts of different reasons ( recently had to 2fa a transaction at a god damn teller window; you just took my DL ).

Isn't the DL (which has a picture) and your face the two factors? Isn't that the whole point of having a picture on a DL?

No, the teller is not sufficiently qualified to be liable enough to match the picture on the ID to the person in front.

The 2nd factor is the phone number on file, which offloads liability for errors in that mechanism to the phone company.

The goal is to reduce the amount of decisions the teller makes, so as to reduce the amount of errors they can make, which also reduces the amount of training they need, all of which reduces costs.

It’s really interesting how the lack of US federal government stepping in to provide an official electronic identity verification API has resulted in the mobile phone networks becoming the de facto arbiters of identity. Even for government services.

I don’t even think I could trust having my phone number on someone else’s mobile phone plan, as I would want to ensure I have as much control over it as possible.

Yes, but it is rather pointless to argue with teller who can't even begin to understand policy dictating it, much less, apparently, make exceptions. Machine told me to do it.

> IDscan was never cryptographic proof you were who you said you were. Maybe better than “enter your name and SSN”

With the benefit of hindsight, we'd have all been better off if SSNs had been so obviously flagrantly public that nobody would ever consider them a trust-factor.

To be fair many SS cards were printed stating right on them they aren't for ID usage.

Yep. Social Security numbers are not guaranteed to be unique, and for a very long time, were pretty easy to figure out if you knew roughly when a person was born and where.

"I know! We'll use this number that its issuing agency says is most definitely not to be used for identification purposes for identification purposes!" is real PHB thinking.

Right, I don't blame the SS administration, they knew what people shouldn't do and warned everyone about it... the problem was private companies were desperate for security-theater and were never punished as they built an entire grand opera-house of insecurity.

What should the companies have used instead? Even the government identifies you by SSN, eg taxes or voter registration. They combine that with state ID often, but the only national ID is passport which not everyone has.

Right, but somehow a bunch of companies still made them into a "prove who you are" factor, and (if we assume incompetence instead of malice) that means a bunch of decision makers still considered the a trust factor.

In some states, until the early 1990s, SSNs were also used as drivers license numbers.

Passports seem a lot better. You scan it with NFC, and the chip inside proves authenticity via asymmetric crypto.

What even would be the fix for this? 153m people need new license asap and id verification systems need to block the stolen ones? Also what are some of the bad things this could cause: a risk malicious actors open verified accounts in their name, ability to vote and travel under stolen id, what else?

3d splatting takes on a different meaning in this context.

Yeah, I mean the last 3 or four big releases of models have come with big scary news. I have to think there is at least some intentional marketing effort behind all this

Or, models are potentially dangerous and it works out in the favor of regulatory capture.

I imagine it like the bumper stickers that say "legalize recreational plutonium"

A maxim I’ve arrived at working w llm every day is “put deterministic things in code, non deterministic in llm”. I do wish the harnesses would be more helpful in this regard. For example I’ve seen tons of scheduled jobs that people wrote in Claude/copilot/etc that could easily have been scripts. They aren’t scripts because the author doesn’t know how to script and they stop at the point the llm does what they want. It isn’t hard to then tell the llm “make a script to do all of the tool calls this prompt needs and then pass the result back to the llm for this non-deterministic part”.

The way I try to illustrate this to my peers, in the context of automating with llms, is to "do as much of the deterministic work as possible before and after involving an agent". Tbf this is largely a restatement of your comment;

Ie instead of

"hey agent, use the api to get X, then do unpredictable mutation Y to it, and email it to me",

it should be;

"Scheduled task runs code to get X then place it in workspace. Task then starts agent in workspace and is prompted to do mutation Y to it. Post-flight code then sends final product."

The difference here being that the pre/post flight stuff is called deterministically, rather than called by an llm.

An absurd example is how harnesses like Codex, Claude code have no 'cut/paste this range of chars' affordance so if you want to move a block of code the LLM galaxy brain rewrites it in token space in the new file.

It ~can~ be done using standard bash utils but there just isn't a kind of standard pattern for 'move this text block via CLI tools' so the LLM and human both overlook it

Worse, the content subtly mutates during that transformation. It's really rare to see that happen with anything affecting execution, but I see comments lost/created all the time.

Anecdotally, when I see coding agents preform this action- I see them using bash. IMO less tools is better, if the agent has a shell- so not having a dedicated cut/paste tool is good.

Preach!

Dedicated Infrastructure > automation > ad hoc scripts > non deterministic llms from a dependability perspective alone. Using an LLM to do something that prettier could have done is analogous to using a 3d printer to nail up a fence.

I’d expect the llm to write the script and run it.

The tooling can remain generic and token spending focused.

What is the “middlest” Mac one could get for this? I’m in the market but keep going back and forth between a 64gb m5 pro or “lower end m5 air and screw it I’ll just pay for cloud tokens”. At current prices the 2-3k diff to try to run something local that isn’t as powerful could buy a lot of tokens.

You can run a lot of these on e.g. M1 max 64gb

Yup. I’ve been evaluating several on openrouter and find token cost meaningless for my work. I haven’t found a great alternative, though the “cost per task” he uses makes some sense.

I’ve been considering a move to local llm setup, having been underwhelmed coat vs value of various online offerings. But at the same time worried anything I get will be obsolete in a couple months. And I don’t want to have to babysit it. I really want some agents managing and creating side hustles for me and have some other things. I’m technical-have written my own harness and use gh copilot and grok daily and have a hosted openwebui+openrouter thing. I’m also torn between a 128g MacBook Pro or a framework, or spark or similar and lightweight laptop to access. Would love advice anyone has for (or against) going local. I have asked ai but have analysis paralysis as 5k would be a big investment for me so I want to make right choices

Well, if you are making side-hustle money now using online models that, critically, you could also run at home, then it sounds like it’s just a matter of numbers. Oh and, unless you spend a lot more than 5k, your local model will still be slower than the online model. What’s your estimated ROI?

Assuming that’s not true based on your phrasing, you’d be shooting yourself in the foot. Start using online models with the same quant at least benchmark as what you could run at home. Prepare for the at home model to be slower.

My thought process is that I don’t mind a slower model if it can work in background for me 24/7 fleshing out side gig ideas I have floating around but no time to focus on myself. I take your point though, and it’s why I haven’t bit the bullet yet. I could buy a lot of tokens for 5k. If I could make that effective then the roi of offline should be something I can calculate fairly easily.

no one is making money side-hustling ai models. This is like reddit wet dream. get real, dont get scammed by ppl selling you these dreams.

Mac, DGX Spark, and a Framework Desktop / Ryzen AI Max 395 (ie Strix Halo) will not give you great performance running LLMs. One benefit of the Spark over the others is you can easily link up to 4 of them. Only MoE (sparse) models will be usable. Even if you can run some massive models, they will crawl. You're better off running one or more GPU cards.

You probably want to try renting some time on a dedicated box with roughly the specs you’re considering and running the open models for a bit to see if you would actually use them before dropping a lot on local hardware. A 128 gig MacBook Pro isn’t going to get you an amazing model, and certainly not amazing speed. GLM 5.2 wants something like 350+ gigs at fp4 iirc.

I think they're just suggesting renting as a way to test that the hardware they're considering purchasing would actually be able to do what they need.

> I think they're just suggesting renting as a way to test

Well, yes, I understood that.

Which is why I started with the words "You don't even need to go that far.".

To re-phrase what I said in clearer terms:

Instead of renting an instance, then messing around with configuring Linux and whatever via SSH or Ansible or whatever. Just point a Hugging Face link at this magic service and get a ready-to-go API back. Enabling you to test your desired model spec with minimum fuss.

Ultimately the guy wants his own hardware. So why waste time messing around with someone else's VM if you just want to test a specific model spec. That is the TL;DR.

Half of my point was to test the models, the other half was to try to get a sense of what the speed would be. Hard to do, but dropping $5k on a 128 gig machine thinking that will unlock good local AI and then realizing that you’ll need to spend >$20k more to run a decent model, and then finding out that even that gives you crap speed isn’t the best way to discover all this.

I very much want local AI to win this in the end, but it’s extremely expensive to run good models at good speed locally right now. Minimax M2.5/2.7, Qwen 3.6, etc are pretty good for basic stuff, but pretty far off from competing with Opus/Fable.

I ran glm 5.2 on rented 8x h200 it could only do 2x concurrency at a cost of $40 an hour. It felt great but dang I wish it was cheaper... It needs 750 at fp8

what was the concurrency limitation? that node should be able to support a lot more

We’ve been using aikido code scanning and pen test tools and been pretty impressed. Will have to take a look at this.

I'm building a competing product and am curious if you'd be up for a conversation about what you've enjoyed best about Aikido and, importantly, what gaps are still not covered.

I have thought that extremely accessible, portable, non-radiation imaging would be revolutionary. Imagine every doctor - or even every person - had a handheld thing they could wave over your affected area to get a high res 3d image of the issue. Of course we have mri, xray, and ultrasound, but those are big and expensive. Obviously there are engineering reasons that is the case, but It seems like a concerted effort to make imaging more portable and accessible, coupled with ai to help analyze could bring about all sorts of follow-on health improvements. Your regular practitioner could check you for heart blockages, clots that lead to stroke, cancers etc as a matter of course. I’m not sure stepping into a golden vat of water is what I have in mind but medical imaging does seem like a possible area for drastic innovation.

Actually doctors tend to frown on people getting full body MRI scans with no symptoms because they tend to cause many false positives.These false positives often trigger a stressful cascade of unnecessary follow-up imaging, bloodwork, and invasive biopsies.