usernametaken29

Now what about those millions of lines of code? Wasn’t the agent swarm supposed to take care of it? All jokes aside this is exactly what happens as you build larger monoliths, system complexity and bug surface increases logarithmically… so in all seriousness you want to reduce LOC as much as possible. Maybe this can teach OpenAI a lesson to produce software that’s correct and adhere to established engineering standards over sensationalised “we built so much code” promotional material.

> As in our solution to the Navier–Stokes problem

Now hold on a minute here.

Our proclaimed, unverified, digressing, stolen manuscript that we presented as the solution to the Navier-Stokes problem.

No worries, I fixed that one for you. No hallucinations needed.

Humans literally need each other to survive. We bend the truth to conform to what we think is convenient and I believe almost everyone knows this. It’s a feature, not a bug. It allows us to have large functioning societies and rampant social injustice at the same time. You can make up your mind about the morality of it for all you want, but it’s really not controversial.

If you use LLMs to digest and summarise software maybe it can help to improve formalisation. In contrast if you use LLMs to produce more code this is contra-dictionary. More codes increases the bug surface circa logarithmically and correspondingly makes formalisation much harder as well. I want to see LLMs be used to strip away code. Only then will I believe they’re really actually good at what they’re doing.

> Part of this shift is the natural expansion of business functions as the company scales.

Every R&D company on the planet has this cycle. Hire lots of developers and researchers. Build product market fit. Lay off all but essential personnel, while increasing the amount of sales and operations stuff as much as possible.

Anthropic this isn’t novel or unique in any way. Stop pretending your AI wonder machines are the reason why. This analysis lacks ANY comparison to other non AI research company hiring cycles. Do your homework please, this is pretty embarrassing.

No

When you want to find partial derivatives for a non linear function you need the Hessian matrix as a guarantee that the term itself will be derivable. But in order to find values to put into the Hessian we often rely on heuristics, otherwise you could have the whole trajectory not making progress or swinging wildly. That’s called conditioning. Due to the nature of the underlying geometry some functions are easier and faster to do this for (scalar multiplication comes to mind) than others (this is why soft max is great because it can be rewritten to fit neatly into the Hessian). But of course limiting the families of functions we choose also means limiting the system as a whole, and eliminating a few very non convex cases from our objective altogether … theoretically reducing the capacity of the models

I wish they’d invest more time in performance improvements over code mods. I recently switched to pi which uses an order of magnitude less RAM (hundreds vs dozens) and noticeably less CPU on my machine… OpenCode is really nice but it’s about as heavy as antigravity and is substandard in features…

Shush now, don’t question our value proposition

This. pi needs about 20MB of ram and Claude Code needs at least several hundred MB if not gigs. How? You’re supposedly magnitudes better at what you’re doing but you can’t get your core product offering (API, billing, accounts, CLI coding) to work properly? The thing people pay you absurd money for???

> Private for all

Literally the first thing that I thought opening this was: “oh great even more spyware in my home” haha. Sure, let’s mount every centimetre with cameras and smart TVs. Even if we buy the premise that there’s no bad intentions, negligence, incompetence and malignant actors can use every iot device as another vector to bug you…

Hell yeah I’m down to become a turtle. Lay by the beach. Eat kelp. Awesome

I can archive my data just fine by using an access based storage tiering like S3 - have your cake and eat it too. Gets around the headache of actually querying all of the 1B+ records because realistically there’s very few use cases for that. If you ever really need it, spin up Kafka and run through the entire archive.

How’s this any different then Iceberg?