My problem with V4 flash is output limit. When I need to write or rewrite a larger file (~1000 lines of code) it will fail with message like output limit reached.
They've both been insisting that humanity's survival is contingent on regulation since the start and employ armies of lobbyists to try to shape the policies they want that hold significant sway in their own country. But of course Americans doing this is smart and entrepreneurial...
Not sure if they support multiple OPDS catalogs now.
I am on v10 (using JF for music only) and with book plugin (forgot which one) it didn't offer to split books into multiple catalogs, which was a bit annoying. Maybe I missed it in v11. Solution for me was to vibecode a simplest OPDS server with 0 css and basic auth.
I often do hand write SVG icons. I know roughly what I want, it's less messy compared to using an editor (cleaner, smaller xml, easier to hand-edit later if needed). Path arc is my nemesis, otherwise it's not that hard. Pelican would take some time, but same as software development, you split it into smaller chunks and do one at the time.
Main problem in complex icon is remembering which (x, y) point is used in which element, <g> with background grid is helpful here. I was even thinking about making extended SVG language with variables for (x, y) points.
I am using SQLite as document db for a side project for years now. Made a custom repository base class that can also store blobs in separate columns, so this type of data is not part of the json document. Today there is also jsonb [1], as far as I remember all functions work the same for json and jsonb.
Also the repo class stores write and delete timestamps as separate columns so I can have CDC. CDC is used for building cached view models and is pushed to object storage every 5 minutes for backup as NDJSON. Another process on home server is restoring the db every couple of minutes for second backup and ready to use DB in case it's needed.
I know there are things like Litestream, I wanted something in process and something that can send alerts on failed backups.
If I create a view of a table containing JSON data (or any other kind of data), I can create columns from calculations on existing fields.
I'm guessing the advantage of generated columns is that they are real columns, not computed columns like in a view. This means that if an insert doesn't work with the new generated column an error will be generated? Is that a correct understanding ?
Is this perhaps the main advantage of this feature ?
The data creation and storage options, always computed or stored on write of dependant columns seems like a possible advantage too.
Main practical difference is that you can put index on generated column, and on view you cannot, SQLite has no materialized views. Constraint part you understood right, NOT NULL on generated column fails at insert, and note that VIRTUAL costs nothing on disk but can still be indexed, so STORED is mostly for when expression itself is expensive.
One of the things that is great about htmx (besides being the best library) is the lack of modern silicon valley corpo communication. It has the slightly unhinged feel of a nerd that spent too much time on Newgrounds or 4chan or something awful. Which is how the internet aught to be.
Z.AI is the only provider for GLM 5.3 on OpenRouter. I don't see 5.3 on Hugging Face. Not sure if this new model is "full GLM" or something smaller, or if they will like Moonshot AI publish weights but put restrictive license [1], which will again leave Z.AI as single GLM model provider on OpenRouter.
I am happy if they publish under restrictive license. Developing model costs a tons of money and company need to make money somehow by still open sourcing project.
As I understand OAI is offering discount only for users using the model via OR. Not sure why. Maybe they want OR users to try the model and switch to OAI subscription or something.
It might also be worth noting that the discount is only if OpenAI itself is used as the provider via OpenRouter. The discount does not exist for Azure or Amazon hosted Sol.
I posted elsewhere, but the Azure uptime & performance for Sol is truly dire. OpenAI is offering 5x faster latency, 4x faster tokens generation, and vastly better uptime (Azure US has only 87% uptime), all for 50% of the price now. I assume the pricing is to compete with other shiny new models (Grok, Qwen etc), but it might also be to cut-off a truly poorly performing Microsoft hosting experience.
Sounds like MoE but more pluggable. Not sure if anyone is researching something like this. I still think your swift or GIS model will need basic reasoning and coding to work, so it's more like multiple smaller models which you can load as needed, e.g. sub-agents for GIS needs the GIS model.
Another approach would be to have basic coding and reasoning model and then load specification for language and libraries into context, it could work for self-hosted models, but I don't want whole specification of the language to be send to API and waste tokens on that.
My motivation is somehow the model should be able to work without hitting an external resource every single time and somehow have all the knowledge necessary locally so we don't get rate limited but how can you stay up to date with the latest information while still remaining a good boy scout by not reaching out to tool use and scraping every single GitHub repo and issue every single time?
I would, once it is possible, specifically unplug the shitcoder expert that sometimes shows up and puts `as unknown as` and bypasses the commit hook that is supposed to enforce it. That should be somewhere near java coder expert in the whatever dimensional space. I can feel it being true.
Could it be that they're using coding agents to develop applications they previously wouldn't spent time on developing? Like internal tools, or experimental builds.
reply