Awesome finally meeting @thruflo IRL for the first time.
We first spoke a few years back when he started @ElectricSQL (which is now part of Neon/Databricks!) but had never actually met!
Excited for all the stuff you're working on now!
And hope to see more of you in London.
@kylemathews 1000s of lines of React context were ruining my life. I watched videos about local first and realized my frontend had become a database. So after a year of building my app, I rewrote the entire frontend with TanStack DB. The big reaction of my cofounders: "The app works now" haha
late night experiment with Vercel's new UI SDK, Tanstack DB, React Compiler.
seems fast enough, kinda like Linear.
might turn this into a template or something
Have to share this awesome open source project: It is called Evil Buttons, inspired by Evil Charts by @legionsdev.
A whole library of crazy buttons. It even supports GPUI.
The creator, @radiumcoders, is only 16 years old. Go give him a follow, he's cooking 🔥
Very clever! @MubitLabs plugs into your @neondatabase AI gateway and trains small, task-specific models for you.
You can then route common tasks to them, saving a whole bunch of tokens/money.
HT to @Shankha_SD and @notanilp for the awesome tech.
The @MubitLabs team built an agent for cutting frontier-model token spend
They tried it on an example template, and it cut frontier-model tokens by 73% and ran 40% faster, without losing accuracy
Running on Lakebase Postgres, Neon Functions, and AI Gateway
Introducing Jerry!
- Your personal desktop explorer and friend.
- Give him missions on your computer and watch him tackle them.
- He keeps a fun journal from his missions.
- JOL (Jerry On Line) gives you remote access with live updates via @neondatabase realtime.
There are two unclimbable mountains in database world: 1 Combined transactional/analytical processing 2 true multi-master scale out
We are making significant progress on both in Lakebase
postgres might seem like a strange choice for a streaming database.
if you compare postgres directly to something like kafka, in terms of pure write throughput, you're going to do about 2x worse, because you pay the penalty of the WAL on postgres. on my local SSD, optimized 64kb inserts are about 300 MB/s on Postgres, and about 600 MB/s on Kafka (with ack waits).
but maximizing disk throughput is very unlikely to be your biggest problem. your biggest problem is infra component bloat. it's the number 1 problem we try to avoid with the open source version of Hatchet. the only strict dependency should be Postgres, and we're going to do everything we can to optimize the shit out of it until we can't optimize it anymore.
so we built durable streams on Postgres, so users don't need to add any additional infra components to their self-hosted Hatchet stack to benefit from durable streams.
We’re extremely excited to announce that Xata is joining ClickHouse!
Blog post: https://t.co/UVGdvOlmVE
And the most important details are here in thread.
signal is a great word for it. Recently I encouraged folks to stop thinking in terms of "loops" and more in terms of:
"forward pressure" - what causes work to start (crashes, bugs reports, performance issues, +100 things)
and
"backpressure" - how does an agent hill climb against an outcome (linters, tests, ai review/risk judge, canaries, feature flags, +100 things)
looping these toward a desired state rather than an outcome / artifact is a powerful mental model - forward pressure increases autonomy in what gets attention, backpressure increases autonomy in getting to a super high-quality solution
9 months ago, loops like ralph were primarily used to go from point A to point B in terms of functionality, but this is just one degenerate case of a more generic control loop:
- read desired state (e.g. no open bugs, or DB perf above X metric, OR "feature Y shipped", OR "all linter warnings resolved or explicity ignored")
- read state of the world (query systems, run linters, etc)
- decide if anything needs to be done (compare to policy / desired state)
(optional: approve decision)
- do the thing (code change, infra change, docs change, process change)
(optional: approval action)
- repeat
shouts out @0xblacklight for the original framing here at AIE back in june. same logic works for evolving the control loop itself
TanStack AI + Neon 💚
Streaming chat, tool calling, structured extraction and summaries, with models from Neon AI Gateway.
One command to get started:
neon functions new tanstack-ai
Go build something cool ⚔️
You can now deploy apps from @cursor_ai.
When you publish an app, Cursor provisions Lakebase Postgres on Neon and deploys it on @vercel. No account to create.