Sam Willis

3.6K posts

Sam Willis banner
Sam Willis

Sam Willis

@samwillis

Building Agents & composable sync primitives @ElectricSQL. Sync Engines, Durable Streams & AI. Started https://t.co/pDO5k5EX42, co-creator of @tan_stack DB.

London UK Katılım Mayıs 2007
794 Takip Edilen3.7K Takipçiler
Sabitlenmiş Tweet
Sam Willis
Sam Willis@samwillis·
My overriding thought going into 2026: Problems you thought were out of reach no longer are. Pick a problem 100x harder than you thought possible to tackle alone, and run with it. Start now. AI will catch up and make it possible.
English
2
5
47
7.3K
Sam Willis
Sam Willis@samwillis·
The last couple of weeks with Fable & Sol continues to confirm this. The discourse that we are "slop" factories is wrong. Slop is a stepping stone to the finished product, it's a POC, it's the underpainting on the finished picture. You're probably not being ambitious enough!
Sam Willis@samwillis

My overriding thought going into 2026: Problems you thought were out of reach no longer are. Pick a problem 100x harder than you thought possible to tackle alone, and run with it. Start now. AI will catch up and make it possible.

English
0
0
2
509
Sam Willis
Sam Willis@samwillis·
Three weeks ago PGlite crossed 10 million weekly downloads. Today it just crossed 11 million! It grew 10% in three weeks 🤯
Sam Willis tweet media
English
9
7
131
11K
Andreas Kling
Andreas Kling@awesomekling·
If you're a married programmer and your wife isn't using at least one of your custom apps daily, you're missing out on one of the most rewarding experiences in programming. Build wifeware!! 🤓
English
120
150
3.7K
145.7K
Sam Willis
Sam Willis@samwillis·
@awesomekling Just showed this to her, and she said I "git commit, but don't wife commit" 😬
English
1
0
3
143
Sam Willis
Sam Willis@samwillis·
@awesomekling Back during Covid we started racing each other doing a sudoku over a morning coffee. It was so obvious to turn it into an app - we played it every day together for a couple of years. Slipped out of the habit, need to change that...
English
1
0
3
1.2K
Sam Willis retweetledi
Oliver Medhurst
Oliver Medhurst@CanadaHonk·
Porffor is now self-hosted, meaning it compiles itself with itself! Not only is this a huge compiler milestone generally, it is also great for dogfooding and also unlocks some fun things further in the future. This is made possible by Porffor's new rewrite, reducing its lines of code by >50% and making C output up to 5x more efficient! Here you can see the stats before and after the rewrite for Porffor itself:
Oliver Medhurst tweet media
Oliver Medhurst@CanadaHonk

good week for some announcements I think

English
11
25
174
17.7K
boris
boris@boristane·
sandbox founders are the new database founders
English
12
4
136
14.6K
Sam Willis
Sam Willis@samwillis·
@BenjDicken I think it's very doable, but it misses the key thing that makes Postgres attractive - the extension ecosystem. I believe it's more compelling to take Postgres as a base and fix its issues. I think swapping parts to Rust can also make a lot of sense, but doing it incrementally.
English
0
0
10
427
Ben Dicken
Ben Dicken@BenjDicken·
I'm curious what it would take a talented engineer to build Postgres from scratch in 2026. Not a slopfork. Written in Rust, from the ground up, and fixing the mistakes of the original design (process per connection, table bloat, schema catalog). Who wants to burn some tokens?
English
194
22
1.3K
169.6K
Sam Willis
Sam Willis@samwillis·
I need GPT5.6 to launch on Monday...
Sam Willis tweet media
English
2
0
13
1.3K
Sam Willis
Sam Willis@samwillis·
What thinking level are you using Fable at? I'm beginning to think that setting it to "Low" is most optimal. It moves much faster and creates a better feedback loop, without really loosing any intelligence. Kind of wish there was an easy way to combine with GPT5.5 though.
English
3
1
6
767
Sam Willis retweetledi
Carlos Baquero
Carlos Baquero@xmal·
Can we merge latent spaces as we merge CRDT replicas? It works for two replicas with a CRDT-like design, and we can do federated querying without exposing the query to the LLMs that process the source data. Abstract and link in thread.
Carlos Baquero tweet media
English
1
2
8
1.4K