daxfohl

23 ms·

331.
by daxfohl 1y ago
Or maybe when you buy a phone you can pay $5 extra to get the OS build that allows sideloading, or make it cost $5 and require you to hand-sign a bunch of forms to upgrade an existing phone to a sideloading-capable version. A little extra f
332.
by daxfohl 1y ago
AI will soon be more addictive and dangerous than drugs.
333.
by daxfohl 1y ago
There's no such thing as a temporary read in LLM land though. Once it's in context, you have to assume everything else connected to the agent will be able to exfiltrate it until the agent is killed and the context wiped. Note, thi
334.
by daxfohl 1y ago
Yeah maybe if an agent workflow is decomposed into steps, each step having certain permissions, and the context optionally wiped or reset back to some checkpoint between steps to prevent accidental leak. This is actually pretty nice because
335.
by daxfohl 1y ago
The tax prep example is safe(r) because presumably it only works with APIs of registered financial services. IDK that a VM adds much. And you can't really block IO on a useful tax service anyway, so it's somewhat a moot example. T
336.
by daxfohl 1y ago
Of course the scary thing is, you're not in control of this. Every company where you've set up an account in the last hundred years is now playing with adding AI features. It's probably only a matter of time before your passw
337.
by daxfohl 1y ago
Yeah, you basically have to think of an agent as malicious, that it will do everything in its power to exfiltrate everything you give it access to, delete or encrypt your hard drive, change all your passwords, drain your bank accounts, etc.
338.
by daxfohl 1y ago
But what if the expansion of the universe is due to some banach-tarski process?
339.
by daxfohl 1y ago
Sure, but do some due diligence. I just say that because I've seen a couple cases where someone does a hack week project that introduces some new approach that "makes things so much cleaner". But then after spending a couple
340.
by daxfohl 1y ago
I gauge it as "the simplest thing to transition". Most of the time, it's easier to transition a single service that doesn't rely on a big number of complex abstractions or extra infrastructure, even if it's at the e
341.
by daxfohl 1y ago
Though in my previous job, a huge amount of complexity was due to failed, abandoned, or incomplete attempts to refactor/improve systems, and I frequently wondered, if such things had been disallowed, how much simpler the systems we inh
342.
by daxfohl 1y ago
I wholeheartedly agree with this. The challenge is perception though. Many managers will see a simple solution to a complex problem and dock you for not doing real engineering, whereas a huge convoluted mess to solve a simple problem (or no
343.
by daxfohl 1y ago
"It’s fun to decouple your service into two pieces so they can be scaled independently (I have seen this happen maybe ten times, and I have seen them actually be usefully scaled independently maybe once)." Same, or reliability-tie
344.
by daxfohl 1y ago
Embed a local redis or sqlite instance?
345.
by daxfohl 1y ago
Boss->VP: "We need to fire people because AI" VP->Public: "We'll replace all our engineers with AI in two years" Boss->VP: "I mean we need to fire VPs because AI" VP->Public: "Replacing pe
346.
by daxfohl 1y ago
Very cool. I've been thinking for a while that this is where things will end up. While custom AI integrations per service/product/whatever can be better and more efficient, there's always going to be stuff that doesn&#x2
347.
by daxfohl 1y ago
Yeah, I think the "pro-linkedin" comments here are probably valid, with the caveat that eventually everyone will quit using linkedin if there isn't more substance on these things at some point. The way it's headed, it fe
348.
by daxfohl 1y ago
Yeah, actually it looks like you're right about the tax code impacts. Doesn't seem like it'd be causing layoffs just yet though, but could definitely be putting hiring on pause due to the uncertainty.
349.
by daxfohl 1y ago
And old unicorns like airbnb and uber now having to compete with traditional hotels and taxis again. I think Elon's takeover of twitter set something of a precedent too: if he could reduce headcount as much as he did and still have a f
350.
by daxfohl 1y ago
A while ago I tried something similar but tried to boil it down to the simplest thing I could come up with. I ended up making a standard maze into a first-person perspective where it unfolds one step at a time, and seeing if a model could s
351.
by daxfohl 1y ago
"When a metric becomes a target, it ceases to be a useful metric"
352.
by daxfohl 1y ago
Okay, yeah my response is ultimately based on the one conversation about it that I've had with the one prof of the one math class I've taken in the last 30 years, so take that with a grain of salt. (Tangentially, I'm so so so
353.
by daxfohl 1y ago
And even aside from defining your own axioms, you can have the same experience just by taking the wrong path in a proof. You'll frequently try something and the goal checker will say that the only thing left is to prove 1 < 0 or suc
354.
by daxfohl 1y ago
I guess it makes sense: the tactics are defined in the lean repo (but not in the kernel of that repo), while mathlib has its own repo. But any changes to tactic code triggers a test to verify that mathlib still works against it. Which impli
355.
by daxfohl 1y ago
The first set is the rules of logic, which applies in non math stuff too. From there you can add new axioms for boolean logic, for integers, for sets, for geometry, whatever. Even given that separation (which itself is fuzzy), even the firs
356.
by daxfohl 1y ago
The surprising thing to me was that tactics are all written in "user-level" code, outside the proof kernel. This makes sense in the sense that you want a small, thoroughly tested kernel that doesn't change. But it also implie
357.
by daxfohl 1y ago
I wouldn't put much money on any of them. IME most mathematicians aren't that interested in formalization, and the gulf between a hand written proof and and a computer verified syntax is pretty huge. Theyre interesting to learn an
358.
by daxfohl 1y ago
None at all. Hit the natural numbers game that was referenced and you can start proving basic things like 2+2=4, n+0=n on up to associativity, commutativity, etc of basic operations, etc.
359.
by daxfohl 1y ago
Wouldn't the be reacting differently if the cost and speed of development in the economy's highest ROI industries was really expected to improve by an OOM in the near future?
360.
by daxfohl 1y ago
I think the files have to be somewhere, just not on the desktop. Maybe root folder or, is it possible to make a folder named ".."? The first step called "move" and the subsequent "read" showed no files, only fo

More ›