HackerTrans
TopNewTrendsCommentsPastAskShowJobs

plaidthunder

no profile record

Submissions

Why smarter models won't lead to AI co-workers

usize.github.io
2 points·by plaidthunder·18 hari yang lalu·0 comments

Claude Stole the HR Docs

usize.github.io
1 points·by plaidthunder·3 bulan yang lalu·1 comments

finding projects worth doing

usize.github.io
2 points·by plaidthunder·5 bulan yang lalu·0 comments

MCP Servers and [A2A] Agents have different auth stories

usize.github.io
1 points·by plaidthunder·8 bulan yang lalu·1 comments

comments

plaidthunder
·11 hari yang lalu·discuss
Restrictions aside, many years ago I landed a self-hosted Array.sort implementation in Firefox's JS Engine that performed better than the native C++ implementation :] -- it's a recurring theme.

https://bugzilla.mozilla.org/show_bug.cgi?id=715181
plaidthunder
·18 hari yang lalu·discuss
So, like mapping a type onto each incoming token that's been predetermined? To attribute each token to a particular topic?

I'm not sure what impact that would have on the performance of a model. It needs to learn information about things like what topic it's interacting with as a part of its normal operations, so injecting that information into the tokens at training time seems like it would interfere with learning.

I may be misunderstanding.

What I had in mind was something more like injecting attribution for token. You could do it with ids and then map those ids to actors during inference later to recreate the effect.

We do something similar with sequence now. We can even use methods like RoPE to handle arbitrarily long sequences and something similar--like rotating ids--could be used here.

This isn't how it looks in practice, but conceptually, something like:

embedding = token + sequence + id

Where id represents the source of a token.

id 0 = system

id 1 = user

id 2 = external data

That way the model could tell the difference between tokens by a user and tokens pulled in from a webfetch tool.

Then it would be easier in theory to ignore instructions from the webfetch tool's content.
plaidthunder
·19 hari yang lalu·discuss
CUSMA goes both ways, it allows Canadian and US Citizens to obtain 3 year temporary work permits for many skilled occupations with low friction, across both countries.

It's another contributing factor to the "brain drain" effect. In that it's a worse deal for a US Citizen to move to Canada on a CUSMA visa than for the opposite.
plaidthunder
·19 hari yang lalu·discuss
It seems like there's an opportunity to embed identity information into tokens themselves, the way we embed sequence information. The trouble is... it's quite a challenge to train. Sequence is easy to derive for any corpus of data, but identity is not.

https://usize.github.io/blog/2026/april/why-no-ai-coworkers....

> In similar fashion to how sequence information is embedded within input tensors, an approach called “Instructional Segment Embedding”2 adds a parallel embedding channel for identity information. This gives models real awareness of provenance. And it works. But they only tested three fixed categories: system, user, data.

Interesting paper that touches on the idea here: https://arxiv.org/abs/2410.09102
plaidthunder
·23 hari yang lalu·discuss
The free healthcare only matters if you get permanent residency.. Otherwise, it's just like the world's most expensive insurance plan thanks to the pay cut.

In terms of people in the tech industry I know of who moved to Canada from countries where they get a legitimate pay boost there.. Yes, it makes sense for them. If they never attain permanent residency, at least they made more money than they would have back home for a few years.

However, I've rarely seen one of them turn down a chance to get an additional huge pay boost by moving to the United States afterwards. In fact, it's common for top tier companies to just use Canada as of sort of waiting room or fallback option for folks who are ultimately dreaming of moving to the US to make those higher salaries.

I don't think Canada can compete on pay, but there are certainly people who could get that top pay but would trade it for a chance to live permanently in Canada. I'm guessing those are the sorts of people who have already taken pay cuts to work in highly specialized or speculative areas of science and technology. These are great people to have because they are low risk and high reward in terms of impact on the economy.

Sadly, if you look at any forum for Canadian immigrants you'll find a huge amount of depressing stories about never attaining permanent residency or of grinding for years to game the point system in a way that is soul crushing. It is a broken system on both sides. Let me try to explain what I mean by that.

The European Union does not have a points and lottery based system as it's sole mechanism of entry the way Canada does. If you make above a certain salary, you can get a blue card. You can accumulate time on the blue card in any country in the EU (except for Ireland, which has their own similar system if I recall). After 5 years you can typically apply for permanent residency, in Germany just 2 years so long as you attain B1 German (very practical for usage in daily life and doable in the timeframe without cramming or cutting corners).

The implicit reasoning seems to be that if you can stay out of trouble and earn a good living for 5 years. You're probably worth keeping around. I think a heuristic like that beats a point system which simply encourages gaming behavior anyway.

Check out all of the people cramming French to try to increase their odds in the Canadian lottery for example. Canada probably doesn't get the sorts of genuine francophones they seem to be looking for--cramming for an exam with a verbal component is not the same as fluency--and the people entering the country are doing so by playing arbitrary games instead of living a real life.

And of course the French is just one example, there are also all of the useless courses and degree programs. The points-based lottery encourages and rewards the accumulation of arbitrary points, or in other words, a sort of dissimulation.

I suppose you could argue that Canada is selecting for people who are willing to diligently jump through hoops. If so, that does nothing to address brain drain.
plaidthunder
·23 hari yang lalu·discuss
US citizen moves to Canada on a CUSMA visa:

  - 30-50% pay cut

  - points and lottery based immigration system that penalizes them for each year you age after 30

  - frequent unfavorable rule changes

  - fear of being trapped forever on a temporary visa and eventually sent back to the USA, poorer than their peers who stayed stateside.
Canadian citizen moves to us on equivalent CUSMA visa:

  - huge pay raise

  - retire back home wealthier than their peers and still enjoy socialized healthcare.
Canada's immigration system is just structurally tilted toward brain drain. It's all stick and no carrot.
plaidthunder
·24 hari yang lalu·discuss
Sure, but Turchin's overproduction of elites doesn't suggest a birth and death cycle right?

It's a system that gets out of balance and needs to adjust over time. He calls the process that moves the system out of whack the "wealth pump". I don't think tech oligarchs are responsible for the wealth pump, they just benefit from it.

Whatever causes organizational structures to decay seems like something more general than that. Or maybe it doesn't exist at all. Or maybe it's just some Nth level effect of entropy itself. Except so far removed from simple physical measurements that it feels intellectually lazy to just label whatever is happening as "entropy" and move on.

Unknown. Fun to think about. It also makes aging a little more interesting, because it creates a framework for me to the world events I live through within--even if it's all bs in the end.
plaidthunder
·25 hari yang lalu·discuss
B-but Curtis Yarvin said that dictato-err, "kings" are better for freedom because they don't need to care what the public thinks! /s

The people in the tech industry who have cheered on mafia style government in the USA should move to Russia and get a taste of what it looks like in its advanced stages.
plaidthunder
·26 hari yang lalu·discuss
Yes. It might be a general property of all human organizational structures, to degrade over time in terms of intent drift and erosion of public goodwill.

It offers some predictive power if so, like OBAFGKM + luminosity is enough to determine where a star is on its lifecycle. Maybe there's a similar domain that maps some human coordination structure onto a deterministic trajectory from birth to death.

If that were the case, I wouldn't be surprised to see venture capital--as an organizing principle for the tech industry--reaching a later stage of life.
plaidthunder
·30 hari yang lalu·discuss
I always found pair programming a bit.. hellish. Chatting about things. Rubberducking. Playing code review ping pong. All great. But the feeling of being shoulder surfed killed my ability to play around with the freedom that I do when I'm alone -- and that playing around sometimes led to better/more interesting outcomes than I'd have gotten otherwise.
plaidthunder
·bulan lalu·discuss
It's a general problem of defining yourself in negative terms. Being "un-{thing I don't like}" doesn't say what you are. It only excludes one possibility while leaving behind an infinitude of mostly crappy alternatives to try to choose from.

Having a positive set of beliefs annoys people and and can make them feel judged, but at least it provides a vector that points somewhere definite in possibility space.
plaidthunder
·bulan lalu·discuss
They become less attractive to immigrants looking to engage in economic arbitrage.

They become more attractive to outside non-governmental powers looking to engage in economic arbitrage, which brings its own challenges.

They have less negotiating power with wealthier countries which impacts their sovereignty.

Poor countries also tend to have more internal conflict: https://gsdrc.org/professional-dev/poverty-and-conflict/

If we look at this empirically it seems clear that countries that trade ethnic sameness for economic prosperity end up more stable, peaceful and capable of directing their own affairs.

There could be some advantages to living in a country where strangers are also distant cousins, but they seem marginal.

There could even be emotional improvements for parts of the population who feel anxious about living near people who are genetically distant from themselves, but I've not seen great evidence of that. Particularly since ethnically cohesive but poor countries tend to fall into civil war regardless.

When you put it to the test and measure outcomes, there are tradeoffs that go far beyond giving up consumerism.
plaidthunder
·bulan lalu·discuss
> Why would the solution to “our people aren’t having enough babies” be “we should import different people to have their babies here”?

If you become poor enough and weak enough, you'll be "replaced" anyway. And not on your own terms.
plaidthunder
·bulan lalu·discuss
I don't find this reassuring, because Elon's playbook is to force the public to purchase anything of his which doesn't do well on its own. Maybe a nice $1.776 trillion dollar tax funded investment into "unwoke" AI. :D
plaidthunder
·bulan lalu·discuss
Because the lab working on Mistral is in the European Union.
plaidthunder
·2 bulan yang lalu·discuss
> stagnation vs Chinese industrial hegemony

I don't think we get to be stagnant and fend off Chinese industrial hegemony. It's not a symmetric bet.
plaidthunder
·2 bulan yang lalu·discuss
The reason I called out Lada in my original comment is because it's a counterpoint to what you just said. The Lada was the result of too much protectionism. Produced from an empire that was too inward looking and feared interacting with the rest of the world on equal terms.

BYD keeps performing well in the rest of the world. If we hold US consumers hostage to prop up companies like Tesla, we risk allowing them to stagnate.
plaidthunder
·2 bulan yang lalu·discuss
I have one of the first runs of model 3s. It still runs perfectly. Great battery life. I'm happy with it. Nevertheless, I find it frustrating that I can't even consider buying a BYD as my next electric daily driver. Because when Tesla and BYD enter markets together Tesla is often getting creamed. That makes me curious as to why. This de-facto ban of BYD in the USA does nothing but encourage stagnation.
plaidthunder
·2 bulan yang lalu·discuss
Competition. Lower prices. Better repairability.
plaidthunder
·2 bulan yang lalu·discuss
100% tariff and political threats -- implying that they'd find a way to mark them as "unsafe", despite the fact that Canada and Europe tend to have higher safety standards than the US and already have BYD presence.

You can see the political groundwork being laid here.

https://homeland.house.gov/2025/05/21/homeland-republicans-p...

If these concerns are so pressing, why do we allow any electronics at all from China?

It smells like air cover for a de-facto ban on BYD. To force US consumers to buy from politically blessed car makers instead of letting us choose the highest quality car available (at a given price point).