Larger programs (like existing Polkadot WASM runtimes) perform a bit worse, but I'm sure there is plenty of opportunities to further detect and optimise specific Rust->WASM patterns to make it more on-par. https://t.co/NEXr7uozZ9
https://t.co/01O3OmbkWV did not have an easy start. Initially I was banned by Anthropic and the first, PoC design (direct WASM -> PVM transpilation) was very inefficient and had correctness bugs. Agents were stuck at being able to debug it and fix. 🧵
I was surprised to see that simple WAT programs have negligible overhead of PVM bytecode 1.5x. Similar overhead is seen between compiling the same Rust program to WASM and PVM.
Note PVM (or rather SPI) has OVERLY INFLATED memory layout declaration which I omit here.
Coding agents are actually, what made as-lan possible. AssemblyScript only compiles to WASM and JAM needs PVM. Few months ago I instructed AI to build WASM->PVM re-compiler. I wasn't hoping for much, but surprisingly https://t.co/01O3OmbkWV produces decent results!
1y ago I had a crazy idea to let people write JAM services in JavaScript-like lang. A balance to steep-learning-curve Rust that is prevalent in Polkadot. Today, let me present as-lan: https://t.co/JAK2o6Ontq Let's iterate fast, since JAM service patterns are yet TBD.
1y ago I had a crazy idea to let people write JAM services in JavaScript-like lang. A balance to steep-learning-curve Rust that is prevalent in Polkadot. Today, let me present as-lan: https://t.co/JAK2o6Ontq Let's iterate fast, since JAM service patterns are yet TBD.
Instead of large, know-it-all models most of us just need fast and good-enough tool-using agents. Not an AI researcher, but my wild guess is that this is disjoint.
That's what THEY want you to think. Looking at the rate of progress of open models and the realistic requirements from an average user I suspect we will rather have abundance of fast, good-enough inference. Just like we have abundance of cpu and memory on our personal devices.
AI KYC is here.
New claude subscribers asked for gov ID & photo.
Not even a regulatory requirement - Anthropic just doing it because they want to.
But regulatory is coming
Next up will be laws:
No AI without gov-issued ID
All AI use tracked to individual - no private AI
Terence Tao proposes what he calls a "Copernican view of intelligence".
Instead of buying into the common, one-dimensional narrative that artificial intelligence will simply evolve from "subhuman" to "superhuman" and ultimately make humanity entirely redundant, Tao urges us to look at the bigger picture.
Much like the Copernican revolution proved the Earth is not the center of the universe, Tao suggests we need to realize that human intelligence isn't the only, or necessarily the highest, form of intellect. Historically, we have treated other forms of storing or creating knowledge—like animals, books, and computers—as secondary. However, we actually exist within a much richer universe of intelligence.
Both human intelligence and computer intelligence possess their own distinct strengths and weaknesses. The true potential lies not in viewing them as direct competitors, but rather in focusing on collaboration. By working together, humans and computers can achieve additional things that neither could accomplish on their own, requiring us to think in much wider terms than just what humans or computers can do alone.
@NirDiamantAI Peter Steinberger told me that he wants PR to be "prompt request". His agents are perfectly capable of implementing most ideas, so there is no need to take your idea, expand it into a vibe coded mess using free tier ChatGPT and send that as a PR, which is now most PRs.
this isn't a huge deal but this is really the flavor of our times
everything is just sloppy. everything has 20% margin of error. nothing has precision
just automate it. just select all. just make ai figure it out
i do it too and it all adds up to a gross feeling world
/btw
Yesterday we held a wonderful cozy meetup dedicated to vibe coding with Claude Code in Wrocław, at the co-vibing space by @fluffylabs_dev. Thanks to @tomusdrw
7 people showed up from different AI verticals (finance, medicine, consumer apps). We vibed a bit together, and there’s a strong desire to scale these events
So, whoever’s reading this from @AnthropicAI — I’m open to serious involvement as an ambassador for @claudeai
Every story has its continuation. My ambassadorship for @claudeai without official ambassadorship already evolving.
We are preparing a series of two events in the coming days in Wrocław.
The first event will be in a cozy format, around 12 to 15 people, to seed the buzz and activate word of mouth. Later we will host a larger event with speakers and panels for 40 to 50 people.
Wrocław folks, stay tuned. The venue is already set - a nice, friendly spot (in the photo). Thanks @tomusdrw
How I build: by voice chatting w/ @ClaudeAI
What I build: 3 projects simultaneously
Who my team: 13 agents on 3 VPSs
How I promote: post and stream on X (haven’t had much time to add that to the pipeline yet. still day one learning @AnthropicAI Claude Code and mastering my skills)
What’s my growth: compared to last week I’ve already freed up some time to search, read and post on X (wanna give a read to my saved posts)
What’s next: one more project (for local community, dedicated to languages 🇺🇦 🇵🇱). Yes, I’m serious about being a Claude Code ambassador.