Joined November 2010
- Tweets 6,304
- Following 565
- Followers 3,119
- Likes 7,497
1,564 Photos and videos
Pinned Tweet
3 Jan 2025
I'm turning 55 this year.
I spent my early years as a piece of shit human. Drinking, smoking, drugs, and wasting my life.
17 life lessons at 55 I wish I knew at 25: 🧵
12
2
57
4,025
Katana's stake page, redesigned.
Cleaner layout, clearer numbers, same productive KAT.
Don't let you $KAT sit idle.
app.katana.network/stake
37
29
268
1,144,597
May 3
I didn't think Jira could disappoint me any further, but they really went out of their way to prove me wrong with the Jira MCP server.
When I die, I want the Jira team to carry my casket so they can let me down one last time.
1
2
198
will button retweeted
Apr 28
Agglayer's pessimistic proof enforces security through one simple rule: a chain can only withdraw what's been deposited. Verified by onchain ZK proofs.
That's how the Open Money Stack's crosschain infra kept @Katana liquid during the rsETH exploit.
Apr 28
DeFi needs to recognize that Finance runs on trust, not software. Liquidity is a posture, not a number. Katana was among the most liquid venues during the crunch created from the rsETH exploit.
Wrote about this and more here: beincrypto.com/polygon-aggla…
28
26
160
17,874
Katana Perps site will be offline: April 27 at 14:00 UTC, estimated ~1 hour downtime.
Open positions will auto-close at upgrade time if you don't close your position beforehand.
No need to withdraw, funds stay put.
Just reopen your positions once the site is back online.
Apr 26
Katana Perps will be offline for a scheduled upgrade
April 27 at 14:00 UTC ~1 hour downtime.
Users do not need to migrate funds. Any open positions will be auto-closed and need to be re-opened after the upgrade is complete.
Join Discord for updates: discord.gg/katananetwork
11
4
57
8,146
will button retweeted
Apr 19
Our team has been actively monitoring the rsETH exploit: Polygon Chain, Agglayer, and the broader ecosystem, including Katana & Vaultbridge, remain unaffected by the incident.
Polygon has safely moved over $2T to date and we will continue to stay vigilant as this event unfolds.
Out of an abundance of caution, the OFT path on Vaultbridge (secured by a 2/3 DVN) has been temporarily paused.
Bridging vbAssets to and from Katana remains fully available via Agglayer.
Agglayer verifies with ZK proofs, not Proof of Authority. Math, not multisigs
40
45
357
33,895
Out of an abundance of caution, the OFT path on Vaultbridge (secured by a 2/3 DVN) has been temporarily paused.
Bridging vbAssets to and from Katana remains fully available via Agglayer.
Agglayer verifies with ZK proofs, not Proof of Authority. Math, not multisigs
14
22
155
67,418
will button retweeted
Apr 15
Fun fact. Before doing dev work, I ran an IT security business for years. While nothing is 100%, I found that human error was the number one cause of hacks.
Layers of security helps a lot.
Isolating critical items from day to day tasks.
Two computer gang 🤙
7
5
32
3,044
Apr 15
Running multiple agents in parallel isn't the hard problem.
Knowing how to break the work down so agents can execute it is.
This is a leadership problem.
Agents opening PRs to merge code isn't the hard problem.
Knowing the code they wrote is the right code for the problem is.
This is a technical problem.
Solve one without the other and you've got nothing.
3
114
Apr 14
The ability to criticise AI will be more valuable than the ability to produce code
3
193
Apr 13
Watching big companies blame AI for layoffs this year has been something.
I spend my days shipping with AI agents. Here's what I see: AI isn't replacing engineers. It's letting the engineers you already have finally tackle the backlog that's been growing for years.
Every team I know has a list of "we'd do this if we had the capacity." AI is the capacity.
If your response to that is "great, now we need fewer engineers" — you're managing the wrong problem.
4
102
Apr 5
Asked a buddy to offer feedback on a new site
Sends back a 5 minute audio file...
Dropped the audio file in NotebookLM -> had it convert each issue raised into a task with issue, steps to reproduce, and expected behavior
Gave it to claude code -> all fixed in < 1 hour
2
140
will button retweeted
Apr 3
Security Notice: Email Impersonation
A bad actor is impersonating Katana team members using domain we do NOT own, email@katana[.]foundation
All official comms from our team are from email@katana.network. Stay vigilant.
6
12
37
6,277
Apr 1
I ran an entire engineering project through AI agents. Let me tell you about the parts that sucked.
1/3
1
97
Apr 1
1. The PRD interview should’ve been a specialized skill. Raw conversation worked for a simple project but would’ve failed to gather requirements for anything more complex.
2. Agentic Amnesia As A Service. Every agent needs a profile and task description that stands on its own. Without this you have a team with great work ethic that doesn’t know shit.
3. 20 parallel agents sounds great until they edit the same file. Coordination is the hard problem, same as human teams. Except humans eventually learn to check Slack first.
4. One-shot solutions don’t exist. Iterate.
2/3
1
64
Apr 1
The hard parts of agent-first development are the same hard parts of engineering management. Communication, coordination, quality standards. The easy parts are the ones that used to be hard: implementation speed, parallelism, coverage.
3/3
38
Mar 26
The interesting part of ClawesomeSingles was different agents for different jobs. Same as a human team, except hiring takes 30 seconds and nobody negotiates equity.
Frontend agent -> UI with actual design quality, not AI slop garbage
ChromeDevTools MCP agent -> opened pages in a real browser, verified rendering
Code review agent -> checks every PR before merge
None used the same prompt. Different tools, different contexts, same codebase, connected through CLAUDE.md
Agent-first development isn't one agent doing everything. It's the right agent for the job.
214