I'm using Kimi K3 + OpenCode. I pay their API pricing, costs about $5 / hour (and chews through ~10 million tokens / hour) during continuous use when I have one or two sessions running and doing their thing.
Can't comment on how it compares to plans (I really don't like the limitations and general shenanigans I see around plans, so I've never tried them).
It is notably slower than Fable / Opus / Gemini, but also vastly cheaper than their API pricing.
Very much agreed. Even more basic than that - memory access patterns are important. The amusing thing is that you end up writing GPU-style code even for CPU. For example - instead of an array of objects, using parquet-style object of arrays is one such trick.
What are some good tools for _authoring_ HDR photography? Lightroom still seems to be mostly SDR. The new Resolve will probably have the tooling, but I haven't tried it.
Huh. When I rented a Z7 ii about 5 years ago, I found their Android app to be pretty great. (My next big camera is likely to be a Nikon, in part due to the nice app)
I remember a story on HN from a while back. The idea is that the larger the org, the simpler the message and the tool has to be to reach everyone. The comment author was saying that as a junior, his company implemented a "tokenmaxxing" scheme for A/B testing - more tests, better for performance review. He, back then, thought it was stupid. However, it got the desired outcome of everyone being familiar with what experiments are and how to run them.
Not necessarily. I'm a proponent of (admittedly not very popular) methodology of "train, do interpretability analysis, adjust model architecture".
It's not more popular for a few reasons: 1) you first need to train a full general model anyhow 2) interpretability is nontrivial and not guaranteed 3) once you make the architectural changes, you can't commit to that architecture as you might miss out in the future with more advancements 4) with modern transformers, there is limited amount of architectural "play" happening.
IIRC the original author of the Lottery Ticket Hypothesis now disavows that idea.
One intuitive way of looking at it is like so - let's say that you have a gaussian-looking plot. You want to fit a gaussian. You have a stupid simple model where you can slide your gaussian left and right.
If your initial starting point happens to be roughly within range, great, your optimizer will take care of it for you and slide it into the correct place. If you're too far, too bad, no meaningful gradient.
Instead, neural nets give you the option to spawn a gaussian anywhere you please. In this case, no sliding is necessary, but it comes at a heavy parametrization cost.
I've personally had a line of thought where you bake in the role into the token. Basically have an embedding (same dim as token dim) for each role, add it to each token. This adds an unambiguous, unspoofable tag.
I ran this with a tiny Shakespeare model (not representative) and had a freeform embedding for each speaker. I ended up with a neat similarity map between every character. (I don't think the map was very informative for several reasons, but that's outside the scope of a small HN comment)
I think this also stems from ML being more like biology or alchemy and less like math or programming (where you can get down to the first principles, abstractions are rock solid, and non-determinism is limited in scope).
On my (admittedly weird) setup, GPT-5.5 Pro times out.
The reading is off because the thermistor resistance also depends on applied voltage, not just temperature. LLMs couldn't get this even after feeding them multimeter voltage readings, not just ADC readings. They went into guessing much more esoteric things like ADC switched-capacitor input current, burnout-detect current sources or IDACs left enabled, board leakage, leaky cap, etc.
I'm personally heavily testing LLMs on electrical engineering problems. I'm finding that it's not meaningfully better at figuring out what's up than the other models.
To give you an idea - here's a very abridged summary of one sample question (originally a full paragraph): I have a voltage divider with a precision resistor and a thermistor, my voltage reading is off by 17%, where's that coming from. None of the models I tested (including Opus 4.8 and Fable 5) could figure it out.