verdverm

10 ms·

241.
by verdverm 29d ago
It took me until the second half of life to understand how important history actually is, thanks for the share
242.
by verdverm 29d ago
I doubt I will ever get on the train that uses "compiler" with "ai" - we're really stretching the analogies with that one
243.
by verdverm 29d ago
In my experience, I have not found any platform's provided skills worthy of inclusion. The number one reason is that they are often huge and cover the entire surface, or there are a bajillion of them. That these are presented as such s
244.
by verdverm 29d ago
Interest rates are not that high, once you move beyond this century's period of near-zero rates which was never sustainable
245.
by verdverm 29d ago
One of the many reasons leadership needs to change, they are now more aggressive towards exploiting their customers, especially in cloud, Kurain is ruining that platform, but the investors like it
246.
by verdverm 29d ago
In this case, I would imagine two scenarios 1. I have markdown, I want a docx version, that is equivalent to html rendering, markdown's whole point, almost certainly many options out there 2. I have docx, I want to make edits in markdo
247.
by verdverm 29d ago
How hard is it to go the other direction? I've been doing some writing things where they use Docx and I use Markdown (have been since before the LLM craze). Boy is it annoying to go back and forth, will be trying the one direction anyh
248.
by verdverm 29d ago
Is it open source? How do we know it is not doing something nefarious?
249.
by verdverm 29d ago
there's a difference between reading the ai generated summary, almost certainly cached for many questions just like trad search results, and actually clicking and interacting with the Ai, where the suggested billions people having gemi
250.
by verdverm 29d ago
my approach is to not use closed weight models, I'm not trying to solve for how to get people to use gemini, quite the opposite actually
251.
by verdverm 29d ago
Are you watching/waiting on your agents? I care zero about t/sec, quality is far more important that quantity or latency, but they run in the background and I check in from time to time
252.
by verdverm 29d ago
sounds vibes and bias laden
253.
by verdverm 29d ago
> brings up a live chat prompt. They're serving fast AI to billions of users Just because it shows up does not mean it is being used. I only click the feedback button to tell them how much I dislike their Ai summaries. They have rem
254.
by verdverm 29d ago
Distillation does not make a model. There is so much more that goes into creating LLM Ai, traces from better models alone is insufficient. You need a good pre training run as foundation, a great RL reward function to make the other model&#x
255.
by verdverm 29d ago
Big Ai is having a moment of reflection and reckoning with open weights, I wouldn't bet either way on a new Gemma release, 4 was distilled from the Gemini 3 series
256.
by verdverm 29d ago
They promised 3.5 Pro at their next or i/o event, but the rumor is that is never going to be released because it would have been embarrassing. They just started letting their engineers use Claude, so it sounds like things may not be go
257.
by verdverm 29d ago
Bernie is in senile decline, he also told us why we shouldn't care about the nazi tattoo because it was "our" guy. Bernie blew my mind with that one and I will never see him the same again. Now he's out calling for 20 ye
258.
by verdverm 29d ago
Not the people we want deciding what regulations look like for Ai. They are deeply invested and in debt, they have motivations that do not align with societal welfare.
259.
by verdverm 29d ago
My experience is that the models are close enough that, with good harness and context engineering, the open weight models can be better for your daily workflows, than frontier without the special treatment. Basically every point of differen
260.
by verdverm 29d ago
They are building another one also in Tibet with 3D printing and robots. They also have the most supercritical turbines in the world and are building ~7 nuclear reactors each year 60GW is what the US is predicted to install across all sourc
261.
by verdverm 29d ago
I don't think people with EVs will make their car charge available for the network, I certainly wouldn't, will need it for myself, plus my car is rarely plugged in
262.
by verdverm 29d ago
So we can share them with each other? Are we going to pass around memory dumps instead? What about tool choice when working on a specific document format?
263.
by verdverm 29d ago
ai slop listicle
264.
by verdverm 29d ago
There are a lot of accounting shenanigans going on in Big Ai, financial reports are misleading, eg Meta "building" a data center through a shell company and then renting it to themselves. It does wonders for the looks of the books
265.
by verdverm 29d ago
I'm a big fan of the Three Inverse Laws for Ai https://susam.net/inverse-laws-of-robotics.html
266.
by verdverm 29d ago
fyi, https://huggingface.co/thinkingmachines/Inkling https://thinkingmachines.ai/news/introducing-inkling/
267.
by verdverm 29d ago
red team humans do this every day, it's not the discrepancy that is the real issue, it's that they are unreliable and we will never know why it did because it has no intent
268.
by verdverm 29d ago
that it knew and ignored/forgot, sounds pretty typical agentic patterns attention is all you need, but it's never enough
269.
by verdverm 29d ago
I've been asking the same about the open weight models, we're buying our tokens from others now, though I think those people are renting hardware from Google in the end anyway
270.
by verdverm 29d ago
It's a slop factory over there apparently. We were sent the greatest slop deck of all time from their sales team. We now have a :cursed-claude: from a slide where they said "we have access to state of the art models like Claude 3&

More ›