Global Metadata Network - Backend for web3
Joined November 2024
- Tweets173
- Following47
- Followers1.7K
- Likes490
Cadastry retweeted
Only way to really know and control what is going on is go local. We are inches away from local OS models doing most of the work. Big labs are not your friends even though you pay them fortunes.
AMD Senior AI Director confirms Claude has been nerfed. She analyzed Claude's session logs from Janurary to March:
> median thinking dropped from ~2,200 to ~600 chars
> API requests went up 80x from Feb to Mar. less thinking and failed attempts meaning more retries, burning more tokens, and spending more on tokens
> reads-per-edit dropped from 6.6x → 2.0x. model stops researching code before touching it.
> model tried to bail out or ask "should i continue" 173 times in 17 days (0 times before March 8).
> self-contradiction in reasoning ("oh wait, actually...") tripled.
> conventions like CLAUDE.md get ignored because there's less thinking budget to cross-check edits
> 5pm and 7pm PST are the worst hours, late night is significantly better. this means the thinking allocation is most likely GPU-load-sensitive.
Cadastry retweeted
Here is what @SECGov released in their Crypto Guidance published TODAY.
The wait is over.. but is it too late now?
Link: sec.gov/newsroom/press-relea…
Cadastry retweeted
JUST IN: 🇺🇸 SEC Chair Paul Atkins announces that "digital commodities, digital collectables, digital tools and payment stablecoins" are not securities.
Cadastry retweeted
🎙️ NEW SLATECAST with Marcus Levin, co-founder of @officialxyo: DePIN at 10M+ nodes, aiming for billions.
Why a data-native L1 matters, proof of origin for AI, a dual token model that aligns incentives, and real world partners from EV charging to mapping.
Hosted by @akibablade and @NateWhitehill
Cadastry retweeted
It's not only about models, it is also about:
1. Creators earning from agents (payment)
2. Many different agents working in teams
3. Digital identities of agents, tools, user, creators
4. Decentralized agentic hosting
5. Open-source models
6. Privacy & security
cc. @3ASprotocol
Cadastry retweeted
Even a single dev can create a small tool, and, if agents need it, it could be called thousands of times a day across the network. Each use will mean revenue.
As more users rely on agents, those agents rely on modular tools to get work done. A protocol makes it possible for those tools to flow between agents, so builders don’t need to reinvent the wheel, they can plug into each other’s work and be rewarded for it.
3AS is how we get there.
Cadastry retweeted
🚨 3AS in action 👇
1/ Introducing YouTube Researcher + its team.
You give one prompt. The lead agent delegates to:
👉 YouTube Surfer – hunts down videos with titles, links, dates, and creators
👉 Google Workspace Assistant – turns those results into a clean, shareable sheet
Yesterday, @3ASprotocol released first video showing 3AS in action. 3AS uses Cadastry as the core backbone of the system to store and manage digital identities of users, agents, tools and engagement agreements.
3AS brings agentic internet to life and Cadastry powers it.
xitter.cf/3asprotocol/status/195…
We’ve teamed up with @WalrusProtocol to launch the Walrus AI agent and toolset on 3AS.
From token launches to research, Walrus AI agent can:
✅Store files on Walrus mainnet
✅Pass blob URLs between agents
✅Make data management easy
✅Work in a team with other agents
Let’s break it down 🧵
Cadastry retweeted
Yesterday we released first video of @3ASprotocol in action.
It shows orchestration, agents working together, usage of tools on URLs.
Users, agents and tools have their digital identities protected with signatures.
Everything talks to everything.
The dawn of agentic internet
xitter.cf/3asprotocol/status/195…
We’ve teamed up with @WalrusProtocol to launch the Walrus AI agent and toolset on 3AS.
From token launches to research, Walrus AI agent can:
✅Store files on Walrus mainnet
✅Pass blob URLs between agents
✅Make data management easy
✅Work in a team with other agents
Let’s break it down 🧵
