genxy

Stop making excuses for the cloud providers. They build these on purpose, for that purpose, working as intended.

Everyone in this thread should be vibe coding legislation with lean 4. We can at least have utopia for a couple months.

It was fucking on purpose, if we had a functioning government, this is one of things they would have nailed them on.

We always did, the clouds convinced us that overages were the norm. You can blame credit ratings as another vector for big business to screw everyone over. Everything should have been pay in advance with an alternate billing method for overages if you want it.

LeCun is an idiot.

We aren't going to get wiped out by a super intelligent AI, we are going to get wiped out by morons wielding intelligent toddlers with the power of a nation state.

I agree with everything here, but lots of people have magical thinking that CXMT will just instantly erase the run-up in memory prices.

Are you telling me that Intel will corner the market on https://www.intel.com/content/www/us/en/history/virtual-vaul...

Because that would be smart. And they have the packaging tech to add compute to it.

Marimo is amazing and "just python", no tangle/detangle necessary.

I don't let it commit the optimization unless it is larger than 5% and no regressions. If it can combine two optimizations and get above 10%, it is allowed.

Not when you use firefox!

CXMT isn't going to just magically price memory like it is 5 years ago. Maybe 10-20% lower, but not some integer multiple lower.

By goods you mean wealth vertically to the oligarchs, then yes.

You don't need ASML machines to make RAM. BS.

What they need to do is all raise prices for 3 years, dump below cost for 4-6 months and repeat.

As curses go, not a curse! https://github.com/loveOSS/awesome-cobol

Once they get good enough, they won't even ship the models, they will just replace all the jobs themselves. This is where Anthropic, OpenAI and Google are heading, esp Google with GSuite, it already is up in your business doing your business.

All your jobs are belong to them.

That is google trying to keep their 10% worth something.

Easy, all the damn time.

It is great news, it means that C++ can slowly wither as other organisms occupy the same niche.

This video on the device itself (from the readme), not about the robot suit itself.

How I See from Inside Robby — AMT630A Heads-Up Display https://www.youtube.com/watch?v=0eQ94DHOay0

It isn't that it is dead, it is just pushed out really really far. So the bathtub curve turns into, "if it survives birth, it lives forever"

> facet is a library for Rust which provides run-time reflection. By annotating your struct with #[derive(facet::Facet)], you get a SHAPE associated type which can be inspected at runtime.

https://github.com/facet-rs/facet

> The core crates give types a SHAPE associated constant with their kind, layout, fields, documentation, attributes, and type-specific operations. facet-reflect builds, reads, and mutates values through those shapes while preserving their invariants.

> Serialization and integration crates build on that reflection layer: JSON, TOML, YAML, MessagePack, Postcard, ASN.1, XDR, CSV, XML, URL-encoded forms, SQLite, Axum, schema generation, structural diffing, and more.

Side by side comparison with SERDE https://facet.rs/guide/serde/

RIP Robin Williams

Shelley Duvall was amazing in Popeye.

AOC

You really need to work on your rhetorical combat, hn is not for amateurs.

I said "techniques in DSPy", your agentic workflows are already using them. Read the GEPA paper. I'll put the link here and you can scroll down and read it.

https://github.com/stanfordnlp/dspy

Your agentic loops are already doing these things.

It is for execs

Just use the post-one-off-project-to-huggingface-skill.md

Techniques in DSPy are still extremely valuable. Your comment is akin to saying, we don't need feedback because open loop control is so good now. It is until it isn't. I don't need a deep understanding of the underlying theory because my gut reaction is always right, until it isn't.

You should have working programs that demonstrate the features you want them to use, and then the skills. Working programs they can mutate in an RL gym.

Maintenance when you don't have access to an LLM?

How small of a model can complete the operation you described above? If writing a parser still requires a software forge with 2T of vram and a petabyte of training data, then I still see value in PEGs.

Maybe the smaller local models, given a structured grammar can use a PEG to generate a parser.