icelancer

probably my single biggest beef with their client and why i don't use it all that often outside of non-interactive mode

Huh. I never considered this. Hacking TVs with firmware would be a great project for LLMs to fuzz.

I've played both. The GM plays tremendously differently than Stockfish.

Engines - specifically heuristically-driven ones like Stockfish - don't play like a strong GM. They play engine-perfect chess, which isn't how a GM plays with any consistency.

I'm only a decent amateur (1550 USCF) but when I lose to a titled player it's largely explainable in human terms how it happened.

I think the point is that a larger model with far less reasoning time solves the problem just fine. Which of course is the tradeoff: The smaller the model, the more reasoning you need to get decent answers to tough questions.

> I can imagine a curve where for a given level of “intelligence” you either need model size, or inference time (“test time compute”), and can somewhat trade one for the other.

This is exactly what was shown on Luna/Terra/Sol tradeoffs - Luna requires much higher reasoning efforts to approach Terra/Sol on lower reasoning. Which is fine, of course, no complaints - but true.

it does this at a variable rate of frames which you can set - not sure if it is transcribing or natively understanding audio, but I think it's the latter since it is much faster than most transcription models I am aware of

Regardless you are right - I can roll my own.... but why

Crazy it's still the only video understanding endpoint. It's what I use it for and no other model even offers a competitor.

Great point.

The EA group have a ton of parallels to the Scientologists, so all of this is not surprising.

I use Modal for fine tuning and unsloth mostly.

This is what I've been waiting for. We are still using fine-tuned deployments of Qwen3.6-27B with a lot of success but could use a bump in intelligence. Here's hoping.

This is actually a very good question.

The funniest and simplest answer is that you make money off yourself.

I asked an adjacent question to a broker friend who flatly replied: "Not interested in being served a Wells Notice."

Which more or less answered the question for me, even if he was being a bit hyperbolic.

Gets a 100% on Pangram. Stuff is so distracting. Write your own posts, FFS. Or at least pass it through "humanizer" type plugins.

dash dash "--" on a lot of systems and word processors turns it into the em-dash automatically

repo is cloned, AGENTS.md is auto-read into context, the doc says to not allow PR spam. think of it like a soft prompt hook.

This is only true for some parts of the time cost function.

love this adversarial work

Lambda calculus?!

Handing over email metadata, or whatever your interrogator wants from you, will only cause them to shift the goalposts, or find something they want to find in the metadata even if it exonerates you.

There is no reason to cooperate with journalists with a slant.

All of us "olds" had this as a hobby. Or used it regularly.

Just a bunch of weird stuff in this article.

No doubt. Especially for GPU inference at scale. We overclock/overvolt for training and tune way down for inference.

Self-racking lets you rack a bunch of gear you'd never find in VM/dedicated rentals, like consumer parts or older, still very good parts. Overclocking options are available as well if you DIY.

If you need single-threaded performance, colo is really the only way to go anyway.

We have two full racks and we're super happy with them.

> What am I missing?

VRAM capacity given the Cerebras/Groq architecture compared to Nvidia.

In parallel, RAM contracts that Nvidia has negotiated well into the future that other manufacturers have been unable to secure.

Interesting - I'll try some other methods

I generally agree with you. But I put up with it since Notepad++ is good software. It is what it is.

I've had a lot of problems with even 10GbE via Thunderbolt 3/4. Bandwidth-wise it works fine, but latency and jitter are issues. This means that stuff like high-speed cameras that need to be synchronized over Ethernet using Precision Time Protocol (PTP) tend to simply fail with these devices.

Yeah, Notepad++ is known for political messaging in their updates. Taiwan, Ukraine, etc.

Their compaction endpoint is far and away the best in the industry. Claude's has to be dead last.