joshjob42 · HackerTrans

5 hari yang laludiscuss

I mean do a bunch of simulation work and statistics stuff but now jack about website design, so if I ever make a website about any project of mine it'll be vibedcoded or a barebones export of an org file README on github. Websites have little bearing on programming language runtime development, no?

joshjob42

17 hari yang laludiscuss

I think a stronger basis for that is probably the Neomacs project aiming to rewrite the elisp layer and all C code in Rust, incorporating GPU rendering etc along the way, see https://github.com/eval-exec/neomacs

joshjob42

17 hari yang laludiscuss

Or that GNU updates a policy that will very rapidly go from probably net silly/mildly contestedly useful to completely ridiculous in a year or two. Not allowing LLM code will be basically turning down the work of the worlds best programmers running at 50x speed in couple years, and will functionally doom any software project that enforces such a policy.

joshjob42

19 hari yang laludiscuss

Add LuaLaTeX and you're cookin' with gas. For real would be fantastic if we could get more or less the full LaTeX ecosystem readily and rapidly available online and in a huge variety of desktop applications.

joshjob42

2 bulan yang laludiscuss

I really like what Qwen are doing, and a lot of these Chinese labs, but until I can ask their models what happened during the student protests in 1989 or why human rights groups are upset about the Uighurs and the model gives me a straight answer I'm just not able to trust these models with anything of substance.

joshjob42

2 bulan yang laludiscuss

An employee often costs a company 2-3x their salary, so someone making 100k a year, costing 300k/yr, who is made 33% more productive (100k more worth of work to the company) offsets the compute cost.

joshjob42

2 bulan yang laludiscuss

Technically true, but if we're talking about local models, overwhelmingly you're gonna be bandwidth bound. You need about 2 flops per active parameter per token. An M5 chip has what, 150-200GB of bandwidth? But it can easily do something like 16tflops of fp16, so you're talking like 100 flops per byte of bandwidth. Which is just to say that in a batch=1 scenario, ie one user, you're only gonna use a few % of the GPU while you're totally saturated your memory bandwidth. For all practical purposes at the consumer level, take your memory bandwidth, divide by the size of the model, and that gives you the max tok/s throughput you're gonna get.

Even a 5090 has something like 50-60 flops per byte of bandwidth, you just can't saturate the compute without running large batches. (At least at inference, prefill is obviously more compute bound).

joshjob42

3 bulan yang laludiscuss

There's a few major problems with the article. The most obvious is that frontier labs are not charging remotely close to the cost of tokens; afaik most estimate north of 80% profit margins. As a reference, providers are profitably providing Kimi K2.6 for $4/1Mtok out. Is that as good as Opus? No, but it's probably at least Sonnet level, so that's ~4x cheaper than Sonnet while still being profitable to serve on the margin. So you aren't plausibly getting into actual subsidization territory until you're over 5:1 sub to nameplate token costs.

How many tokens can you realistically burn through in one chat session? Opus and many other frontier models do maybe 60tok/s, less 250k/hr out. In you can use more, but in most cases cache is 5-10:1 cheaper than new input. Say you average 500ktok in, 90% cache, per request. That amounts to 100-150ktok in new input-equivalent costs, which in most cases is ~20-30ktok in output-equivalent costs. Do a request every minute, that's a total of about 1.5-2Mtok/hr. At API prices that's $50/hr for Opus, but really it probably only costs Anthropic $10/hr to serve that.

That said, even if a developer is burning $50/hr, many, many employees at large companies cost more than $100k/yr to employ all costs considered, so making them say 20-30% more productive can easily make that worth it for most. If the labs shave their margins ultimately to more like 20-30%, you'd have ~$15/hr in costs to use the services, and nearly every white collar job is way over 30k/yr to employ. If your salary is 80k, you probably cost the company 200k all in, so making you 15% more productive offsets the $15/hr cost.

So first party providers are not in a horrifying position or anything from a subsidization standpoint. The people in bad shape are Cursor and Perplexity, who don't have frontier models and are dependent on the open source community, which is typicly 6-12 months behind the frontier. They have to pay full freight API costs at 80% margin for the big boys to serve their harnesses, which is indeed untenable, and they'll have to either force users to use open source models and/or in house models they can serve at-cost or they will have to charge vastly more.

Gemini, Claude, and ChatGPT first-party services like Antigravity, Codex, and Claude Code are not in serious trouble though.

joshjob42

4 bulan yang laludiscuss

"If you distribute modified code, or offer it as a networked service, you must make the source available under the same terms. This is not a restriction on sharing. It is a condition placed on sharing: if you share, you must share in kind." -- This is, on any plain reading, a restriction on sharing. "You can share only under these conditions" is plainly more restrictive than "sure do whatever you want". You can argue that it's a restriction that ultimately leads to more sharing overall. But it is a restriction on sharing in any given case of sharing nevertheless.

joshjob42

4 bulan yang laludiscuss

If the cost to copying code based on specifications, tests, etc is so close to zero as to be functionally zero cost, then any user can simply turn their AI on any library for which there is documentation and any ability to generate tests, have it reverse engineer it, and release their reverse engineered copy on GitHub for others to use as they like.

So I'm not sure it matters whether a giant company uses it because random users can get the same thing for ~ free anyway.

joshjob42

5 bulan yang laludiscuss

My night and weekend project the last month or so has been creating and implementing a package that provides a pure s-exp syntax for Julia that lowers to Julia's AST directly, and lately been churning through (mostly Opus is doing the actual churning) all the problems of creating an automatic transpiler for Julia to this other syntax.

Not ready to share just yet but nearly at the point that there are no Julia-syntax fallbacks in the entire base/stdlib and a femtolisp parser for the sexp syntax is able to build a complete Julia sysimage from the transpiled files. Already verified that I can transpile the .jl source of the Julia package for the syntax into the syntax, then use that transpiler to transpile again and load into the running sexp repl, then use that transpiler on the source again and get byte identical code, and along the way am testing to ensure that the entire Julia test suite passes in the sysimage being built.

So, with any luck here soon I'll have a sexp syntax for Julia that builds from raw transpiled sexp-syntax source and uses sexp syntax natively in the repl but can transpile & load any Julia code. Fingers crossed.

I'm aware of --lisp but it's not very good imo lol.

joshjob42

5 bulan yang laludiscuss

I really don't see how that is true.

For instance, once you develop atomically precise manufacturing ala Drexler and have a complete model of biology, etc., drive solar panel efficiency to very near the upper theoretical bound for infinitely many junction cells for a raw panel of ~68%, then there isn't really anywhere to go that matters for humans. Material production would be solved, anything you could desire would be manufacturable in minutes to hours, a km^2 of solar panels could power 10-20k people's post-scarcity lives.

You eventually reach the upper bounds on compute efficiency and human upload model efficiency -- unknown but given estimates on upper bound for like rod logic (~1e-34Js/op), reasonably bounds on op speed (100MHz), and low estimates for functional uploading (1e16 flops), you get something in the zone of 0.1nW/upload, or several trillion individuals on 1m^2 of solar panel in space. When you put a simulated Banks Orbital around every star in the Milky Way in a grand sim running on a system of solar panels in space where the entire simulated galaxy has a 15ms ping to any other point in the simulated galaxy, what exactly is this infinite stream of learning? You've pushed technology to the the limits of physical law subject to the constraint of being made of atoms.

Are you envisioning that we'd eventually be doing computation using the entirety of a neutron star or (if they can exist) a quark star? Even then, you eventually hit a wall where physics constrains you from making significant further gains.

There is an ultimate end to the s-curve of technology.

joshjob42

5 bulan yang laludiscuss

Most wouldn't because it's expensive. But at scale automated vehicles should be dramatically less expensive, in the range of 50-60¢/mi conservatively, and at that level it is going to be quite compelling to a lot of people since it's a private vehicle (no taxi driver) and it's reasonably affordable, a 1 seat ride, etc.

It's possible they'll be even cheaper but that range is the cost according to the IRS of operating a typical vehicle all in, and that seems like a reasonable guess of the cost of an autonomous electric vehicle with far lower probability of crash than a human (all the savings basically going to profit margin).

At ~60¢/mi, there'd be a lot of people who would save money on balance using autonomous taxis to get everywhere vs owning a private vehicle (10k mi/yr would cost only ~$6k/yr, a pretty low cost of ownership/use for a private vehicle).

joshjob42

10 bulan yang laludiscuss

Actually it's mostly patch files but they're ignored by github.

joshjob42

10 bulan yang laludiscuss

People leave the Culture, and you can form communities of various sorts, you just can't force people to not leave them. In a world with unlimited abundance, no disease, optional death, and more or less unlimited morphological freedom and the ability to form any consensual voluntary community you like with freedom of exit for all members, what possible change could one be demanding?

joshjob42

10 bulan yang laludiscuss

Maybe it's just me but I do semi-regularly have my phone slip out of my hand and hit me in the face while in bed, haha.

joshjob42

10 bulan yang laludiscuss

My OG Ultra lasts 3 days if I turn off the always on display, which I do because it doesn't serve much of a purpose, I can just tap the watch to wake it. It charges from 0 to full in 1.5 hours, pretty linearly, so dropping it on the charger for half an hour or an hour while I'm on a work call every other day or so keeps it plenty charged.

This one will have even more battery life, and gets 12 hours of use in 15 minutes, which I suspect will mean for me without the always on display I may well be able to charge it only while I'm actively in the shower (when I'd take it off anyway as I hate wet bands) and be good for the day.

joshjob42

10 bulan yang laludiscuss

I don't want/need the whole thing to be flat but I do prefer it to be stable. For instance if the plateau were a bit thicker so that the camera lens was flush with the surface (even just an extra bar sort of inside the plateau) it would mean that when I put it down it would never rock back and forth when I'm tapping at it on a table.

joshjob42

10 bulan yang laludiscuss

But lighter than any iPhone since except maybe an SE.

joshjob42

10 bulan yang laludiscuss

I'm going to preorder one because I want a light phone and a large screen. This will be the lightest iPhone in years while also having a bigger screen than most. I dropped from the Pro Max to the Pro last year because I was tired of how much it hurt when I dropped my phone on my face.

I don't have much call for most of the camera system, and my battery life on my Pro is just fine. I have plenty of chargers typically, and for emergencies or times I know I'm going to be out I could potentially get the battery pack.

I basically never use cases on my iPhone, and at most will maybe use an ultra-thin one or some sort of structure adhered to the plateau just to make it flat across so as to not rock on a table.

comments