HN user
I've data where i cannot store metadata that i need to search semantically so i embed it on the fly at every search with static embedding and tokenizing was more than 99% of the cpu time. Granted that was due the naive implementation of the default tokenizer which was o^2 with document length and just switching to a proper scanner solved most of it without going to simd and whatnot, but still.
Compute will become the means of production and we are not going to get a share of it because the world is hyper optimized for value extraction
these materials are transparent to light it is not a property of the human vision system that light goes trough N2
steering:
You are the sole owner of the project implementation.
User maintained documentation:
- goals.md for the project overall goals
- tech.md for guidance on how to build the project
Agent maintained documentation, current state living specs, these are not logs:
- project.md is a map of the code, components and features.
- choices.md write here all decision taken by the user.
Do not duplicate information between these document.
Same. For some reason late opus model are very superficial doing ux work and so am using gpt for that, but backend is much better engineered by claude, gpt prefer to duplicate everything it needs on the spot causing class sprawl
Very good on vision, really helped oneshotting complex ix thay I had so far to buukd piecemeal. Ended the 200$ plan weekly allowance in two days, so theres thay.
Mybatis was a thing even back then... you still need a domain model after all
never said purely functional, but there are only 4 data channels in each method (input, return, calling another method, setting a state) - and if you constrain your class design to pick only 2 for each method, your life is just a tad easier.
and doesn't matter how bad the rest of the world is, rest of the world is some other maintainer's problem, you just encapsulate it.
Seems hella inefficient.
Better method start to realizing that everything that every program do is data transformations and or movement
Then you ask llm to subdivide data in a tree along the domain model, classifing streaming vs storing nodes
Then for each node you discuss with the ai for the best data structure
Then you ask for an interface that fully encapsulate the structure and every mutation only allows to go from a valid state to a valid state and bidding else is allowed to touch the state
And that's mostly it just connect all the interfaces until input goes to monitor or to storage or to api or wherever the destination is
I really dislike opus 4.8 it rarely compete things and prefer to waste tokens making lists of things that are missing. When stuck or need input it words the challenge at length without conveying anything useful for decision making, and quite often its solution to problems is to excise features or just try catch errors and proceed with faulty data silently
I hate loud ads as well as anyone else and I welcome this resolution but I would not treat the challenges regulation poses as simplisticly as this. There is a lot of research in increasing loudness without increasing decibels, especially for concerts, but it migrated to ads when tv started adding automatic volume controls to normalize across services.
Anthropic is subsidizing their enterprise customers by up to 40 times, and OpenAI up to 70 times
might as well be the other way around with non subscribed token being 50x overpriced, or any combination thereof
also uber was non profitable for the longest time, raking up 31b in losses, on the bet of capturing the market worldwide. scale here is different, but it's also 10 years later, with a lot more volatility and floating cash in the market (voo grew 327% over that period, not unreasonable that round size grew on the same trajectory)
There literally are "no u turn" signage where you are supposed not to do that. They literally put up signs for it. It is not glowing in the sky, and it doesnt need to be, and doesnt help making a point strawmanning it.
Until they reverse on a highway and kill a family. Being able to drive isn't where parent poster put the bar
yeah been away from a while. thanks!
Does BAR have chicken maps?
most merge improve a small subset of "feeling" benchmark (too small, too specific, or out of distribution) and tend to show degradation on actual benchmark, with especially punishing result on long chain benchmarks.
also only work on matching architectures (i.e. finetunes/loras of the same model)
I use glm for all code investigations and top level system design of all kinds, and then present finding to confirm and act upon to opus. everything that burns token goes there.
the finding aren't always accurate, but it saves ton of opus token
likewise I have google ai from my photo storage, so I give claude / opencode a skill that uses gemini (agy now) command line for web searches, using their flash model line.
*Anthropic requires it
Metadata does not require data and conflating them doesn't help your point
The right level of recording of non crime is zero wdym reduced?
Same, even opus favor short term solution and scripts with a billion flags that constabtly require rescanning to understand how to launch it is a constant struggle to get it to build sane default and reusable scripts that run with minimal parameters
But they are, especially those with batteries
Fiat Multipla level design blunder
Also nobody talks enough about the fact that workforce is effectively cut out from the means of productions. Even with the capital at hand blackwell cabinets are all sold out, contracted to the big providers.
There are paralles to the industrial revolution, but it seems the working class is cut out from being able to deny labor in exchange for better conditions.
The demand doesn't necessarily double.
System security is not a human value. Access key rotation effective immediately is a compliance requirement, and completely orthogonal to human decency, which is delivered trough garden leave or severance, not extended system access
https://arxiv.org/pdf/2508.09101
tldr 2% average point lost on Rust compared to python, gap vary by model, go has a better upper bound but opus had it 3% below python.
benchmark is a bit old but research on why is there, article is just vibes
All the control plane. Data plane is distributed and roles using iam to access resources can still do so during a control plane outage.