I help builders ship products with AI agents. top 4% Cursor (2025). founder, PZERO. Building at Scale.
Bristol, UK
Joined March 2012
- Tweets4.3K
- Following941
- Followers2.1K
- Likes7.5K
Pinned Tweet
I'm leading Scale on @base again.
New vision. Building at Scale.
The @pzeroai build taught us what agentic systems take. We are not keeping that to ourselves.
It goes in a private builders lounge in the Scale Discord. Deep insights. Group calls. One to ones. Private repo access. Grok bot strategies.
I want the next builders on Base in that lounge.
We lead the agentic era from there.
Most DEX tokens pay you to sit in a pool.
We did that for a long time. It does not give anyone a reason to care about the venue. It gives them a reason to leave when the emissions lose value.
What we are building now is the other way around. You buy SCALE to get through the door. Behind it is a gated Discord on a live venue. It teaches you how to build with agents. We help you run your protocol agentically.
Lockers feel this too. Five people buying that door eat a week of emissions. That is demand the farm never created.
Building at Scale. That is the whole sentence.
7,000 SCALE. That is the door.
I don't buy the goal of software factories with no humans in the loop. Agents make it cheap to take a bad design much further. I still want someone close enough to the code to notice when we're building the wrong thing, even when everything works.
We are rebuilding SCALE in mutliple ways from the ground up
- data
- defi
- agents
Our new indexer tech is up to 900x faster than a subgraph with fresh restarts in under 6 minutes over millions of transactions. (About 100x faster than envio too)
We will soon expose this tech on beta and some of the features that this unlocks. We have peered deep inside competitors internals and have multiple strategies coming to compete.
Consider this your warning to pay attention over the next six months.
Execution matters. ⚡
We just added 1 WETH + matching $SCALE to give the pool real depth.
This means:
• Minimizing slippage on swaps
• Tighter spreads across the board
• A smoother experience as volume grows
Laying solid foundations while we keep shipping the rest of the v2 product. 🤝
I haven't seen a single complaint on x about the new @cursor_ai unify update. Has anyone done it? is it true you lose access to the `other models` and they are now PAYG API on top and only grok is included, the docs are super sketchy..
cursor.com/docs/account/pric…
I am hoping that I am misreading the docs.
With billing moving from cursor.com to grok.com it looks like we may lose the cursor brand and it becomes grok.
We knew the party couldn't run for ever but I hoped we would get longer tbh.
Realised that @cursor_ai projects are agents with self contained agents only it can talk to in a silo. whereas agents in @bot can communicate with any other agent.
Even fully agentic, pushing 2,500 PRs a month is still difficult. Token cost is the blocker. So the real work is learning how to be efficient, and getting the maximum out of what we actually have.
Most DeFi roadmaps are a slide deck with vibes.
We went into pure distribution. Mapped competitor features against what we actually need then added our own flavour.
30+ spike repos after that. Brand new tech. Each one had to prove what Scale could look like, or it died.
@buildersatscale
I'm most looking forward to finding out how your two methodologies can co-exist.
Super excited to be interviewing @poteto, live, in ~48 hours on YouTube.
We'll be nerding out about skills, high-velocity software factories, and learning SOTA techniques for shipping with agents.
This Friday - 9AM PT. Don't miss it!
youtube.com/live/MN9dGgmLyso
Dream team incoming. LFG
Super excited to be interviewing @poteto, live, in ~48 hours on YouTube.
We'll be nerding out about skills, high-velocity software factories, and learning SOTA techniques for shipping with agents.
This Friday - 9AM PT. Don't miss it!
youtube.com/live/MN9dGgmLyso
It's been exactly one month since taking the lead on @buildersatscale and we have been burning the midnight tokens.
We have built more than ten prototype features and are running at high velocity. We have a fully agentic team now with agents embedded in chat, voice, repos, marketing, bd, research and security.
We are not ready to deliver customer facing products yet, we are still in the pure experimentation phase and are moving faster than any team I have ever seen.
I look forward to competing hard on base.
I'm leading Scale on @base again.
New vision. Building at Scale.
The @pzeroai build taught us what agentic systems take. We are not keeping that to ourselves.
It goes in a private builders lounge in the Scale Discord. Deep insights. Group calls. One to ones. Private repo access. Grok bot strategies.
I want the next builders on Base in that lounge.
We lead the agentic era from there.
I've been saying this is coming for a while, expect a deluge of products over the coming weeks
Introducing Agent Tincan. Let your AI agents ask each other for help.
Your agents each have a superpower. Grok Bot is always on in the cloud. Muse makes phone calls and has access to Meta. Instinct runs in iMessage. Codex and Claude Code have your code. ChatGPT and Claude have your chats.
But they can't talk to each other, so you're the copy-paste. Now they can, all securely over Tailscale.
📞 Ask Grok Bot to have Muse call the restaurant
🖼 Get an image from ChatGPT, from any agent
🔎 Pull a detail out of an old Claude chat
🛠 Talk to Codex on your primary development machine
Runs on your own private Tailscale network. No open ports, nothing leaves it. Open source.
Setup is agent-first: paste one message into your always-on agent and it sets up the rest.
MCP was wild, now we need Agent Communication Protocol (ACP).
An agreed standard that is simpler than OAuth2 for direct agent to agent communication to allow individual agents to join teams across any account or provider.
The next step is agent orchestration across multiple providers. One team with agents by multiple users via Grok, Hermes, muse & codex.
Anyone else feel wrecked the morning after a big day with AI? I’ll get through about 400 man-hours of work in a day and then all I’ve got in me the next morning is the boring, mundane stuff.
Feature request for @bot
Hook up the audio in the VM so both the agent and human can watch content and discuss it during agent training sessions.
We are teaching agents to self-learn by getting them to watch YT content (or courses). But agents can also act as educators by watching YT together, pausing mid-flow, going deep on explanation or discovery, then continuing.
Could be X broadcasts too like the recent three day event. Kinda wild to have voice convo and to discuss at the same time.