HNHacker News
TopNewBestAskShowJobs

cavisne

1,688 karma · joined October 17, 2015

submissionscomments
cavisne··on Requiem for a Solar Plant
Its a good read but are his intentions unambiguously good? He had a windfall from crypto (ie contributing nothing to society). Then he was desperate to avoid paying the tax he owed on that windfall.

Is there any actual need for this Solar Farm in the middle of nowhere (that was only built there because of a tax scheme)? Are Texas ratepayers meant to cover the cost of the interconnect in their $/Kwh instead of him?

Better to just pay his taxes and move on, and leave the subsidies for an actual useful solar project.

EDIT: Oh and mineral rights are basically the original cryptocurrency/memecoin, so its somewhat funny that they came into play

cavisne··on Google is using YouTube videos to train its AI video generator
Hilariously there was a story how Google could not train on Youtube data due to their TOS, so they changed it for new videos. Meanwhile everyone else was scraping Youtube as much as they liked and training on it.
cavisne··on I Read All of Cloudflare's Claude-Generated Commits
Some API's (Gemini at least) run a search on their outputs to see if the model is reciting data from training.

So for direct copies like what you are talking about that would be picked up.

For copying concepts from other libraries, seems like a problem with or without LLM's.

cavisne··on Unfit for Work – The startling rise of disability in America
Nothing compared to the NDIS rort in Australia. There is literal gang warfare over providing "services" to people with high payouts.
cavisne··on Doge Developed Error-Prone AI Tool to "Munch" Veterans Affairs Contracts
Anyone who thinks the VA is well run should watch a few episodes of the “financial audit” YouTube channel. The most financially irresponsible people in the world (and generally bad people in general). Almost all are cashing a “disability” payment from the VA.
cavisne··on The ‘white-collar bloodbath’ is all part of the AI hype machine
My theory for these PM's is its basically a cheap way to take potential entrepreneurs off the market. Its hard to predict if a startup will succeed but one genre of success is having a Type A "fake it till you make it" non technical cofounder who can keep raising long enough to get product market fit.

These types all go to the same schools and do really well, interview the same, and value the prestige of working in big tech. So it's pretty easy to identify them and offer them a great career path and take them off the market.

Technical founders are way trickier to identify as they can be dropouts, interview poorly, not value the prestige etc.

cavisne··on Deepseek R1-0528
"knowing why a model refuses to answer something matters"

The companies that create these models cant answer that question! Models get jailbroken all the time to ignore alignment instructions. The robust refusal logic normally sits on top of the model, ie looking at the responses and flagging anything that they don't want to show to users.

The best tool we have for understanding if a model is refusing to answer a problem or actually doesn't know is mechanistic interp, which you only need the weights for.

This whole debate is weird, even with traditional open source code you cant tell the intent of a programmer, what sources they used to write that code etc.

cavisne··on TeleMessage Explorer: a new open source research tool
Only one person in the groupchat needs to be using Telemessage, ie. a CIA agent can use a government device with Telemessage to talk to sources on Signal. Signal has a great protocol & robust clients, and getting caught with Signal on your phone is probably a bit better than being caught with CIAChat on your phone.

The actual implementation here is atrocious though.

cavisne··on TeleMessage Explorer: a new open source research tool
The signal protocol is public, using their servers is frowned upon but its not a source code license violation.
cavisne··on AlphaEvolve: A Gemini-powered coding agent for designing advanced algorithms
From the paper it was a speedup on the XLA GPU kernel they wrote using Jax, which is probably not SOTA. I don't think Jax even has a official flash attention implementation.
cavisne··on How to Write a Fast Matrix Multiplication from Scratch with Tensor Cores (2024)
"An H100 GPU has 989 TFLOPs of half-precision matrix multiply compute, and ~60 TFLOPs of “everything else”"

I always thought there was a lot of crossover between gaming GPU's & DC GPU's (and the volume is why NVIDIA is so far ahead). Are tensor cores somehow related to the pre tensorcore SM's (like an abstraction on top of SM's?)

cavisne··on CVE program faces swift end after DHS fails to renew contract
MITRE has a trademark on the term CVE.
cavisne··on Tesla Releases Stripped RWD Cybertruck: So Much Worse for Not Much Less Money
Hybrids are gas cars.
cavisne··on Tesla Releases Stripped RWD Cybertruck: So Much Worse for Not Much Less Money
This is somewhat true but there is also a chasm where you earn good money but not enough to afford a mortgage. This chasm grows wider ever year in countries like Canada and Australia (and to a lesser degree America)
cavisne··on Trump exempts phones, computers, chips from ‘reciprocal’ tariffs
So now the finished products are tariff free, but the input costs when bought into America are tariffed. The exact opposite of what you would want to do to bring manufacturing back to America.
cavisne··on Google will let companies run Gemini models in their own data centers
This is being sold as an air gapped product, it has to work offline by definition.

Sure you could hide some way of phoning home and deploy it into the SCIF, but would you really want to risk a firing squad to improve some advertising metrics?

cavisne··on Google will let companies run Gemini models in their own data centers
Seems pretty high, this is an air gapped product so at some point the employees of whatever government they are giving it to would need to SSH into the VM's to load new weights etc. Lots of ways to make it tricky/watermark the weights though.
cavisne··on Google will let companies run Gemini models in their own data centers
The raw computation is just a bunch of matrix multiplications in a row, most of the algorithmic complexity/secret stuff would be around scaling & efficiency.

For training the model the HW is much more important as you need to scale up to as many chips as possible without being bottlenecked by the network.

This would just be inference, and it doesn't need to be very efficient as its for on prem usage not selling API access. So you could strip out any efficiency secrets, and it would probably look like a bigger Gemma (their open source model).

I wonder if they would/could try and strip out stuff like whatever tricks they use for long context + video support (both of which they are a bit ahead of everyone else on).

cavisne··on Germany creates 'super–high-tech ministry' for research, technology, aerospace
Europe is mostly "old money" so this makes sense. They didn't earn that money they are custodians of it.
cavisne··on How a $2k 'Made in the USA' Phone Is Manufactured
Factually despite all the noise there are no significant tariffs in either direction at this point in time, 98% of trade is tariff free under the free trade agreement. Mexico and to a lesser degree Canada are in a more favorable position than they were before the tariffs as they are the only countries in this position [1]

[1] https://semianalysis.com/2025/04/10/tariff-armageddon-gpu-lo...

cavisne··on Pentagon to terminate $5.1B in IT contracts with Accenture, Deloitte
In addition they use their government work to increase private sector profits, either implicitly "we work with your government" or explicitly (PWC offering briefings on tax policy they were helping with to the targets of the tax policy)
cavisne··on Gemini Live with camera and screen sharing capabilities
None of that onboard stuff ended up working afaik as the Tensor chips shipped right as giant LLM's became how dominant.
cavisne··on New White House AI Policies Introduce Government by AI
"In addition to the use of Signal, some DOGE staffers are bypassing other vetting processes and chains of custody for official government documents by working simultaneously out of Google Docs instead of circulating single copies of drafts, a source briefed by a government official said. “There's multiple people in one Google Doc editing things simultaneously,” the source said, referring to the online word processing software. That was partly how DOGE was working so quickly, the source added"

Outrageous!

cavisne··on Gemini Live with camera and screen sharing capabilities
You can do the same from the browser on any phone (including iphone) by going through AI Studio

https://aistudio.google.com/live

cavisne··on Ironwood: The first Google TPU for the age of inference
Was that gamble wrong? I thought all LLM training workloads do collectives that involve all nodes (all-gather, reduce-scatter).
cavisne··on Trump temporarily drops tariffs to 10% for most countries
These complex 4D chess strategies never really pan out. Maybe someone in the administration thinks that was the goal but not Trump.

During the campaign Trump said he wanted 10 percent tariffs on the world and 60 percent tariffs on China. After all the noise we are now at… 10 percent tariffs with the world and 125 percent tariffs on China (pending negotiation with China).

cavisne··on Ironwood: The first Google TPU for the age of inference
The JAX docs have a good explanation for how a TPU works

https://docs.jax.dev/en/latest/pallas/tpu/details.html#what-...

Its not really useful for other workloads (unless your workload looks like a bunch of matrix multiplications).

cavisne··on Ironwood: The first Google TPU for the age of inference
Nvidia has ~60% margins in their datacenter chips. So TPU's have quite a bit of headroom to save google money without being as good as Nvidia GPU's.

No one else has access to anything similar, Amazon is just starting to scale their Trainium chip.

cavisne··on Why Do Domestic Prices Rise with Tarriffs?
"that kind of thing is going to come to America... and be automated". Pretty disingenuous cut on the quote there by the bluesky post.
cavisne··on The Llama 4 herd
The H100 theoretical flops number is just marketing, as it relies on sparsity that LLMs don’t use
← PreviousPage 3 of 23Next →