Pinned Tweet
Built a quick uptime checker with Next.js, and Node on AWS Lambda: onlineornot.com
5
8
146
starting a Workers project from scratch for the first time in ages, wild how much stuff you need to remember to tell your agent to not do its default slop
- effect
- oxlint/oxfmt/dmmulroy/anti-slop
- pnpm
- latest LTS version of node
- hono
- cloudflare.config.ts
- sandbox 1.0
- effect
- oxlint/oxfmt/dmmulroy/anti-slop
- pnpm
- latest LTS version of node
- hono
- cloudflare.config.ts
- sandbox 1.0
22
4
326
15,942
Max Rozen retweeted
I’m generalizing here but it’s amazing how many people running agents like this (24/7) and burning tokens like crazy have nothing of substance to show for it.
substance = things that significantly impact growth in their business
substance = things that significantly impact growth in their business
A genuinely serious question: ——All you saying that you have agents running all night doing things. What are these things that you need to do constantly?
8
3
38
1,746
Max Rozen retweeted
Oh you've got strong beliefs weakly held? That's great. I've got correct beliefs. But your approach sounds fun too.
0
1
19
535
Max Rozen retweeted
Before/After - I know what I am reading
There has been a lot a lot of chatter here about getting agents to speak ASD-STE100 being nice but for observability agents BLUF and SBAR for handovers/status updates are very powerful
en.wikipedia.org/wiki/BLUF_(com…
en.wikipedia.org/wiki/SBAR
8
12
525
48,036
Max Rozen retweeted
This is my Minister for Productivity, dawg. We’re gonna go into a recession.
In a randomised trial, software developers using AI took 19% longer to complete tasks. Yet they believed AI had made them 20% faster.
At the ‘Evidence in Action’ event, I argued for a government that tests what works, and saves money by learning from the results. ministers.treasury.gov.au/ministers/andr… #randomistasCommunity noteThe 19%/20% figures are from a July 2025 METR study of experienced open-source developers using early-2025 AI tools, which the authors describe as an outdated snapshot; a 2026 follow-up found some evidence for speedup with newer tools.
metr.org/blog/2025-07-1…
metr.org/blog/2026-02-2…
arxiv.org/abs/2507.09089
6
9
200
3,532
Max Rozen retweeted
hot take: you don't need a software factory, and i don't love the term. it's just marketing
all you need is good software engineering and investing your tokens in the right places. the more deterministic the better (and token efficient)
all you need is good software engineering and investing your tokens in the right places. the more deterministic the better (and token efficient)
all the companies selling you a software factory produce a shockingly tiny amount of software
439
578
8,082
492,560
Max Rozen retweeted
Building your own agent harness is today’s “build your own irc client”
0
1
6
218
Max Rozen retweeted
Having the ability to straight-up conjure code has helped with speedrunning the coding part.
Ah but coding was a load-bearing strategy to avoid marketing, so...
Ah but coding was a load-bearing strategy to avoid marketing, so...
2
3
12
3,042
