tlarkworthy

Oh they just fixed this 9d ago and I guess this comment provoked them to close the issue!

I would say it doesn't but it is actively working on it

https://github.com/ClickHouse/ClickHouse/issues/52054

I see some BlueSky users mirroring their content from Mastodon

Also trialed Devin, it's quite impressive when it understands the code formatting and local test setup, producing well formatted and test case passing code, but it seems to always add extraneous changes beyond the task that can break other things. And it can't seem to undo those changes if you ask. So everything requires more cleanup. Devin opened my eyes to the power of agentic workflows with closed loop feedback, and the coolness of a slack interface, but I am gonna recommend cancelling it because it's not actually saving time and it's quite expensive.

Thanks you, I have it at last! Been watching from the sidelines for 10 years!

Can someone give me an invite. I am in Europe and I don't know anyone who has an account to vouch for me

I was excited when I read the title but after reading I am at a loss to understand how Netflix is solving the problem. Centralised cost dashboards??!?

How do they solve it on AWS? Tags, Accounts, something else? Why can't they use cost explorer?

you are looking at the wrong app https://observablehq.com/@tomlarkworthy/ink

oh wow, they have the fluid demo too!!! Thats so much better!

https://scrtwpns.com/mixbox/fluids/

very easy, its a free parameter in the simulation see the "Ink dissipation" slider, you can set it to > 1 for some weird effects.

I tried it as a background on Android but it killed my battery within hours so I stopped. Shame, its so cool.

I have a fork of this that inverts the light model from additive to subtractive and suddenly its like ink in water

https://observablehq.com/@tomlarkworthy/ink

read the article, close the feedback loop with something verifiable (e.g. tests)

And then you discovered gimble lock, and to fix that you just need one extra float...

> I can already type faster than I can think.

But can you write tickets faster than you can implement them? I certainly can.

That your JSON deserializer accepted them.

I was recently annoyed to find postgres indexes don't support skipping [1] you also can't have the nul character in a string (\u0000) [2]. Its great, but it has some strange WTF gaps in places.

[1] https://wiki.postgresql.org/wiki/Loose_indexscan

[2] https://stackoverflow.com/questions/28813409/are-null-bytes-...

Isn't this obvious when you work with a stochastic system that giving it tons of wrong moves is gonna increase the failure rate.

I've benchmarked R2 and S3 and S3 is well ahead in terms of latency especially on ListObject requests. I think R2 has come kind of concurrency limit as concurrent ListObject requests seem to to have increase failure rate when serving simultaneous requests

I have a few of the S3-like wired up live over the internet you can try yourself in your browser. Backblaze is surprisingly performant which I did not expect (S3 is still king though)

https://observablehq.com/@tomlarkworthy/mps3-vendor-examples

You can do it without using an append only logs https://github.com/endpointservices/mps3

However it will be much simpler with the new conditional writes

Some things that helped me

Every query response in SQL is a rectangular matrix [1]. JOINs add columns sideways. UNIONs add rows vertically[2].

[1] Which is why tree shaped data is awkward to query in SQL in one round-trip.

[2] From this you realize why the column names have to match to apply a UNION, and why recursion is something to do with a UNION.

there is a culture of type safety which is a very watered down machine checked proof that the thing entering a function has affordances X,Y and Z. Its not the end to end functional verification you are talking about but its not nothing either.

It detects naming collisions, orchestrates recomputation with a topological sort and manages module reuse and normalizes generators and promise to a common abstraction. The parameter indirection is for separating the inner cell code from the inter-cell dependency graph. You need to do that for the partial recomputation to work without the user manually plumbing in the dependencies (like you have to do in React). I will read the TC39 I don't know it so can't compare.

Edit: yes seems like signals could be a major building block to do the recomputation part

> and Observable is cool, but strays from standard JS.

The front end does but the underlying runtime is running just javascript, and the source code is basically javascript with some trivial macros which are fully captured in the MIT licensed acorn parser. That's why normal Javascript debugging expression work perfectly in Observablehq.

https://github.com/observablehq/parser

I love the Observable runtime. I wrote a decompiler for it so you can bidirectionally convert between the front end source and the compiled pure JS representation.

https://observablehq.com/@tomlarkworthy/observablejs-toolcha...

I actually did that in observable userspace https://observablehq.com/@tomlarkworthy/exporter

astro is a modern framework "optimised for content heavy websites" i.e. vanilla assets mostly except when asked for something more.

[1] https://astro.build/

I think we already standardized with the file as the unit of document exchange. You can make a CRDT interface to that file if you want, but CRDT is should not be the authority. There are too many competing implementations and what is best is too context specific and we already have the file as the standard everywhere so it's never gonna catch on

Akka and Erlang both support distributed routing to their actors, but this is planetary scale and fully-managed out of the box, which is very cool.

It's the actor model essentially.

You can have a DO proxy each user connection, then they forward messages to the multipler document. The user proxy deals with ordering and buffering their connection message state in the presence of disconnects, and the document DO handles the shared state.

I wrote my own Notebook Assistent and I use it to quickly make data driven arguments. It is able to both code and read program outputs so it's very good at pulling remote sources, then peeking at the result, doing the data wrangling and then displaying the output. I think having the LLM being able to read program state is a big upgrade for what you can achieve, it can do test driven development for example because it can read the test report. However, being a side project limits the fluidity of the UX, it's open source and buildless though if you want to modify it.its very hackable by design. https://observablehq.com/@tomlarkworthy/robocoop