Sunil Pai
Systems and developer-tools engineer at Cloudflare as of 2026, on durable infrastructure for AI agents; worked on React and PartyKit before that. Writes at sunilpai.dev.
Sunil Pai did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
Latest x.com
What they believe
-
People are still judged for things entirely outside their control, things they were born with and cannot reasonably be expected to change.
it's a sad reality of the world we live in that people are still judged for things entirely outside their control, things they were born with and cannot reasonably be expected to change.
x.comSaid 14 Oct 2026This on korrents
-
Original thoughts do not scale.
also, these original thoughts, they don’t scale. every new one needs an investment of time and energy, and there’s no guarantee that either that one’s good or the next one’s going to be good. you can’t reuse much of the work, and maintenance itself is a huge cost.
sunilpai.devSaid 12 Oct 2026This on korrents
The bold line is korrents’ wording; the quote is theirs. These 2: what they wrote before what they said aloud, one per source, newest first.
Feed
As its own page →Hiding
17 October
16 October
15 October
14 October
-
Their words
it's a sad reality of the world we live in that people are still judged for things entirely outside their control, things they were born with and cannot reasonably be expected to change.
13 October
-
Usestheir ownrcmnd.app
CloudflareTheir words
96% of remaining projects would fit on cloudflare's free tier
12 October
-
korrents.com
Original thoughts do not scale.Their words
also, these original thoughts, they don’t scale. every new one needs an investment of time and energy, and there’s no guarantee that either that one’s good or the next one’s going to be good. you can’t reuse much of the work, and maintenance itself is a huge cost.
11 October
-
Their words
every piece of this stack would’ve been rejected for weekend projects a year ago because you’d have to spend too much time/effort/“innovation tokens” because of the unfamiliarity. now it’s actual fodder for ambitious and innovative ideas
8 October
From one piece Collecting favours 3 beliefs · sunilpai.dev
-
Their words
we made a contraption that speaks back to us of the secrets of the universe and it becomes a moral responsibility to parse it and better the human condition a worthy endeavour and one I hope the species lives up to
-
Their words
you just help people, and then ten years later something happens and all these relationships that you didn’t even realise you were building suddenly become visible.
-
korrents.com
You shouldn't make people earn your help by being perfectly polite while they're having a bad time.Their words
you shouldn’t make people earn your help by being perfectly polite while they’re having a bad time.
+ 1 more
-
korrents.com
When someone has a problem, you should solve the problem first and deal with everything else later.Their words
someone comes to you with a problem, just help them with the problem first. figure out the rest later.
6 October
From one piece @threepointone on X 2 beliefs · x.com
-
korrents.com
In 2026, there will be a cheat code machine that has infinite breadth and depth for pattern matching.Their words
Unfortunately, in 2026, well, you have a cheat code machine that has infinite breadth and depth for pattern matching.
-
Their words
The thing I learned this year is that when engineers say they like working on “hard problems”, they actually don't. What they like is puzzle solving, and having a repertoire of patterns that they can pattern match onto traditionally hard problems, usually involving databases or something with scale, etc.
5 October
-
Their words
I need to shake my public facing friends at AI companies (devrels, posters, etc) and beg them to start using their natural voice more and stop performing like it's the zirp years, it's a red flag and you're painting yourself into a corner.
4 October
-
Their words
what that implies about your information diet / environment, and whether it’s good for you
3 October
-
korrents.com
Consciousness is just a pattern-matching mechanismTheir words
consciousness itself is just a pattern matching mechanism. bugs have consciousness.
-
korrents.com
Machines can be consciousTheir words
I absolutely think machines can be conscious, and it’s not a dramatic thing to believe.
-
korrents.com
Something currently called psychosis should at some point, perhaps now, be recognised as actually being euphoria.Their words
at some point (now?) we should stop calling it psychosis, recognise that it’s actually euphoria
27 September
-
Their words
people think I'm joking when I tell them that to develop taste for software they need to have brunch with their friends, watch old movies and find old records, walk around metropolitan cities and observe how the building styles lend to the personality of the people around them, but I'm so fkin dead serious.
19 September
7 September
1 September
-
Recommendstheir ownrcmnd.app
Next TokenTheir words
you should watch the podcast named after making stuff up all the time : @NextTokenShow
-
Likedtheir ownrcmnd.app
Durable ObjectsTheir words
the best days of durable objects are ahead of us
14 August
12 August
9 August
24 July
20 July
29 June
24 June
21 June
17 June
15 June
28 May
11 April
2 April
1 April
13 March
28 February
23 February
27 January
Nothing matches.
Beliefs
Korrents What they believe 29 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
People are still judged for things entirely outside their control, things they were born with and cannot reasonably be expected to change.
it's a sad reality of the world we live in that people are still judged for things entirely outside their control, things they were born with and cannot reasonably be expected to change.
@threepointone on X Said 14 Oct 2026
Original thoughts do not scale.
also, these original thoughts, they don’t scale. every new one needs an investment of time and energy, and there’s no guarantee that either that one’s good or the next one’s going to be good. you can’t reuse much of the work, and maintenance itself is a huge cost.
original thoughts have terrible UX Said 12 Oct 2026
The current technology stack allows for ambitious and innovative ideas that would have been rejected as weekend projects a year ago due to the effort required by unfamiliarity.
every piece of this stack would’ve been rejected for weekend projects a year ago because you’d have to spend too much time/effort/“innovation tokens” because of the unfamiliarity. now it’s actual fodder for ambitious and innovative ideas
@threepointone on X Said 11 Oct 2026
Show 26 more
When someone has a problem, you should solve the problem first and deal with everything else later.
someone comes to you with a problem, just help them with the problem first. figure out the rest later.
Collecting favours Said 8 Oct 2026
You shouldn't make people earn your help by being perfectly polite while they're having a bad time.
you shouldn’t make people earn your help by being perfectly polite while they’re having a bad time.
Collecting favours Said 8 Oct 2026
Goodwill compounds in ways that can become visible many years later.
you just help people, and then ten years later something happens and all these relationships that you didn’t even realise you were building suddenly become visible.
Collecting favours Said 8 Oct 2026
Parsing the secrets of the universe revealed by contraptions that speak back to us is a moral responsibility and a worthy endeavour.
we made a contraption that speaks back to us of the secrets of the universe and it becomes a moral responsibility to parse it and better the human condition a worthy endeavour and one I hope the species lives up to
@threepointone on X Said 8 Oct 2026
When engineers say they like working on “hard problems”, they actually don't; they like puzzle solving and pattern matching.
The thing I learned this year is that when engineers say they like working on “hard problems”, they actually don't. What they like is puzzle solving, and having a repertoire of patterns that they can pattern match onto traditionally hard problems, usually involving databases or something with scale, etc.
@threepointone on X Said 6 Oct 2026
In 2026, there will be a cheat code machine that has infinite breadth and depth for pattern matching.
Unfortunately, in 2026, well, you have a cheat code machine that has infinite breadth and depth for pattern matching.
@threepointone on X Said 6 Oct 2026
People at AI companies should use their natural voice more instead of performing.
I need to shake my public facing friends at AI companies (devrels, posters, etc) and beg them to start using their natural voice more and stop performing like it's the zirp years, it's a red flag and you're painting yourself into a corner.
@threepointone on X Said 5 Oct 2026
A good work life is about the work, not the environment
what that implies about your information diet / environment, and whether it’s good for you
@threepointone on X Said 4 Oct 2026
Something currently called psychosis should at some point, perhaps now, be recognised as actually being euphoria.
at some point (now?) we should stop calling it psychosis, recognise that it’s actually euphoria
@threepointone on X Said 3 Oct 2026
Machines can be conscious
I absolutely think machines can be conscious, and it’s not a dramatic thing to believe.
@threepointone on X Said 3 Oct 2026
Consciousness is just a pattern-matching mechanism
consciousness itself is just a pattern matching mechanism. bugs have consciousness.
@threepointone on X Said 3 Oct 2026
Developing taste for software requires real-world experiences like brunch with friends, old movies and records, and observing city architecture.
people think I'm joking when I tell them that to develop taste for software they need to have brunch with their friends, watch old movies and find old records, walk around metropolitan cities and observe how the building styles lend to the personality of the people around them, but I'm so fkin dead serious.
@threepointone on X Said 27 Sept 2026
A person's ability to make decisions for many people is questionable if they struggle to understand their own preferences.
mf you can barely decide what discomfort you're subjecting yourself to in finding out what you like/dislike, and I'm supposed to trust you to make decisions for millions of people?
@threepointone on X Said 27 Sept 2026
Bringing AGI to the world requires a specific story of how it will benefit everyone.
You can’t bring AGI to the world without a specific story of how it’s going to benefit everyone
@threepointone on X Said 21 Sept 2026
Big projects are not done with big efforts but with slow, steady, incremental work.
big projects are not done with big efforts. it’s a big marathon, just slow, steady, incremental work.
the senior engineer death spiral Said 19 Sept 2026
Engineers are in the reputation-building business, and software is downstream of that.
remember, you are in the reputation-building business, and software is actually downstream of that.
the senior engineer death spiral Said 19 Sept 2026
New York would be a wonderful city to walk everywhere if people were not openly smoking weed on the streets.
new york would be a wonderful city to walk everywhere, if everyone and their mother wasn't openly smoking weed on the streets
@threepointone on X Said 15 Sept 2026
The Foundation sci-fi series is really about individuals with extreme agency and puzzle-solving ability having wildly outsized impact on a universe that would otherwise sit waiting to decay.
foundation is ostensibly a sci-fi series about empires, psychohistory, and hard technology, but really it's about individuals with extreme agency and puzzle-solving ability having wildly outsized impact on a universe that would otherwise just sit there waiting to decay
@threepointone on X Said 14 Sept 2026
A million experiments are needed in security against autonomous attacks and entities.
we really need 1 million experiments in this space of "security against autonomous attacks and entities"
@threepointone on X Said 9 Sept 2026
Coding agents made implementation cheap without making roadmaps shorter or software better -- the binding constraint was never how fast code gets written.
buuuut my roadmap is’nt ten times shorter. and I’m definitely not ten times better at deciding what is worth building. teams haven’t started casually shipping a year of product work every month. you can look around and and it’s hard to say what software has gotten meaningfully better in the last year.
the task isn't the job Said 14 Aug 2026
Agent autonomy is a capability, not a product direction: what the person is actually trying to do decides whether more of it helps or ruins the thing.
we tend to describe progress in agents as progress in autonomy: first they could do five minutes of work, then an hour, now we talk seriously about agents running for days. that’s impressive, but autonomy is a capability, not a product direction! the Job To Be Done tells you whether to use it.
the task isn't the job Said 14 Aug 2026
A machine can complete the task and completely miss the job -- coding makes this easy to see precisely because its tasks are so legible and verifiable.
a cybernetic machine can nail the task and completely miss the job. coding makes this very very easy to see because the tasks are so legible (and verifiable!): write the function, migrate the API, fix the tests.
the task isn't the job Said 14 Aug 2026
Cheap intelligence does not make agency cheap, and these tools may mostly compound the advantage of people who already have the skill, taste, money and time.
making intelligence cheap does not automatically make “agency” cheap. people are constrained by time, money, confidence, health, education, institutions, family, geography; a model does not magic any of that away. there’s a perfectly believable future where these tools mostly compound the advantage of people who already have the skill, taste, money and the spare time to use them well.
the task isn't the job Said 14 Aug 2026
Organisations say they want disagreement but charge social capital for it, until people stop evaluating the argument and start evaluating the source.
we say we want disagreement, but every disagreement costs the person making it a little social/political capital. do it often enough and you become the negative person. people stop evaluating the argument and start evaluating the source.
every company needs a cassandra Said 9 Aug 2026
Organisational dissent suits an AI precisely because it pays none of the costs -- no promotion to want, no ego, no performance review -- that silence the humans who do it.
an ai is oddly well suited to this job. cassandra doesn’t want a promotion and doesn’t need the people in the channel to like her. disagreeing with the vp three times this month won’t show up in a performance review. she doesn’t have an ego invested in being right either.
every company needs a cassandra Said 9 Aug 2026
An agent told to disagree with everything is worthless: it has to form its own view and speak only when that view differs enough to be worth the interruption.
“you are a devil’s advocate, disagree with everything” is probably the worst possible prompt for this agent. within a week everyone would ignore her. she’d become the slack equivalent of a smoke detector that goes off whenever somebody makes toast.
every company needs a cassandra Said 9 Aug 2026
Beliefs others hold too
A good work life is about the work, not the environment 2 hold this
what that implies about your information diet / environment, and whether it’s good for you
@threepointone on X Said 4 Oct 2026
Machines can be conscious 2 hold this
I absolutely think machines can be conscious, and it’s not a dramatic thing to believe.
@threepointone on X Said 3 Oct 2026
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Recommends rcmnd
rcmnd What they recommend 7 things, 8 mentions — each backed by an exact quote.
Compiled by rcmnd.app, not by them: the quotes are theirs, and each links to the post it was copied from.
Recent
Uses Cloudflare
Loved Cloudflare
Liked Durable Objects
Show 5 more
Recommends Next Token
you should watch the podcast named after making stuff up all the time : @NextTokenShow
x.com 1 Sept 2026
their own
Recommends Real-time sync engines
there’s also a great post on sync engines by Adam Nyberg
sunilpai.dev 16 Feb 2025
Liked Workers AI
Workers AI has dope models for your use. No API keys when used inside Workers.
sunilpai.dev 10 Dec 2024
their own
Uses Cloudflare Workers
Anyway that’s why I’m using Cloudflare Workers for my AI agent stuff. Race you to the finish line.
sunilpai.dev 10 Dec 2024
their own
Uses Cursor
I use the Cursor IDE regularly, and especially for the braindead stuff, I’m mostly hitting tab to autocomplete.
sunilpai.dev 3 Dec 2024
Where they publish
Blog sunilpai.dev Essays on agents, durable execution and developer tools.
Regular posts, several most months, on agent infrastructure, durable objects and what building developer tools is actually like. Titled “Solving the decision problem”. Has a feed.
Recent
- original thoughts have terrible UX 12 Oct 2026 thinking for yourself is a terrible business model.
- Collecting favours 8 Oct 2026 Notes on reputation; A thing I learnt from dad
- Do you actually like hard problems? 6 Oct 2026 I spent years becoming the engineer people reached for. Now they still reach for me, but for different reasons.
Show 17 more
- the senior engineer death spiral 19 Sept 2026 on proving yourself, burning out, and being a good teammate
- the appeal 7 Sept 2026 what does the creator owe you
- the task isn't the job 14 Aug 2026 a new role, and some questions I want to build my way through
- startups need an asymmetry 12 Aug 2026 A quick note on startups
- every company needs a cassandra 9 Aug 2026 an ai agent for the socially expensive work of organizational dissent
- one document, two hands 20 Jul 2026 the agent belongs beside you, not between you and the app
- a letter from the orchestra pit 24 Jun 2026 the work of the musician is not the work of the machine
- never waste a token 15 Jun 2026 durable inference: resumable streams, crash recovery, and why the LLM request shouldn't die with your process.
- Tlön, Uqbar, LLMs, Tertius 28 May 2026 The World Will Be Prompt
- developer relations after the cheat code machine 2 Apr 2026 (corollary- developer relations is becoming more about helping humans develop judgment around these systems)
- let the code do the talking 13 Mar 2026 (why llms and safe sandboxes may change the basic contract between users and software)
- the context is the work (what the day-to-day looks like now) 7 Jan 2026 (corollary- pr descriptions are becoming the apprenticeship surface for remote teams and coding agents)
- where good ideas come from (for coding agents) 3 Jan 2026 (and the part where users have to level up)
- guitars and javascript - the case for jank 14 Apr 2025 worse is better (and that’s why i still write javascript)
- drop that meet link 17 Mar 2025 just keep talking
- an event bus for ai agents 27 Feb 2025 it is very professional yes
- ai agents are local first clients 16 Feb 2025 sync engines finally have a killer app
Link verified 2 Sept 2026. Recent items update automatically from the channel.
x.com @threepointone Constant, and mostly about agents.
High-volume commentary on agents, the web platform and developer tooling — where an argument gets worked out before it becomes a post on sunilpai.dev.
Recent
- we didn’t start the fire 2026 sucks ass 17 Oct 2026
- Air France has starlink AND classic cinema?? 17 Oct 2026
- say it with me now cooooode mooooode 17 Oct 2026
Show 17 more
- 5am pint, SF bound Big week ahead 17 Oct 2026
- The only print media I pay for. It’s so good. 16 Oct 2026
- My completely uneducated guess (intuition?) about mechanistic interpretability is that it's all code. These synthetic emotions: code. Reasoning: code. It’s all code. And if it's code, that means we have decades of knowledge, intuition, strategies, just things that will let us understand it and then modify it. And now that we think that we can compile back and forth from neural networks and, let's call it, “human-readable” code, we will be able to “solve” alignment the same way that we figure out and control determinism in code. Going to follow this intuition for a bit and see where it takes m… 16 Oct 2026
- great website, love it here 16 Oct 2026
- Sobering and distressing. Watch now. 16 Oct 2026
- double rainbow bro 16 Oct 2026
- this was an amazing read. gave me all kinds of energy. 16 Oct 2026
- .@ADoricko hey rainmaker got a shout out on blowmymind! very cool 16 Oct 2026 instagram.com
- flying out tonight to SF, back in a month. adventures! 16 Oct 2026
- codemode solves all of these btw 15 Oct 2026
- new take: now that we're getting past the cope of "design it for humans and it'll be fine for agents" we should revisit all apis made for human comprehension and cut them brutally down to efficient primitives and let the agents figure out what's best for them instead via @jdan, and QTing one of my more popular takes that *might* not hold water anymore 15 Oct 2026
- today's long read sorted 15 Oct 2026
- this is what the first version of this looked like 15 Oct 2026
- tweet of the year right here 15 Oct 2026
- when I see that someone I've never interacted with has blocked me, I have to assume the obvious. it's a sad reality of the world we live in that people are still judged for things entirely outside their control, things they were born with and cannot reasonably be expected to change. I wish we lived in a more accepting society, but unfortunately prejudice runs deep, and sometimes all you can do is acknowledge it and move on. they simply cannot bear to live in a world where someone can be this hot and handsome and smart. I understand. 14 Oct 2026
- holy shit is that an esp32 lmao 13 Oct 2026
- some numbers as of today: - ~ 9k dead projects deleted - ~2.4k emails sent out - 96% of remaining projects would fit on cloudflare's free tier all deploys being stopped on friday, this shutdown's going smoothly. also got a bunch of thank you mails that made me feel things 🥲 13 Oct 2026
Link verified 2 Sept 2026. Recent items update automatically from the channel.
GitHub threepointone Cloudflare's agents SDK, PartyKit, and older React work.
The work itself — Cloudflare's agents SDK and Workers tooling now, PartyKit before it, and React contributions before that.
Recent
- sunilpai-dev — my site 12 Oct 2026
- signal-bloom — A low-latency guitar effects studio with an AI-assisted pedalboard. 24 Jul 2026
- ai-sdk-resume-race-repro 29 Jun 2026
Show 12 more
- agent-recovery-viz — An interactive visualization of Cloudflare Agents' chat recovery system. 21 Jun 2026
- pizzo 21 Jun 2026
- klocker — kloc churn viz 17 Jun 2026
- test-do-rpc-fetch-ordering 11 Apr 2026
- dynamic-workers-playground 1 Apr 2026
- emojico — CLI tool to convert emoji to favicon assets 28 Feb 2026
- agents-sdk-openai-websocket-fetch 23 Feb 2026
- cloudflared-login-worker 27 Jan 2026
- test-zod 21 Dec 2025
- aywson — Modify JSONC while preserving comments and formatting. 20 Dec 2025
- get-biggest-dependents 19 Dec 2025
- repro-vite-exports 18 Dec 2025
Link verified 2 Sept 2026. Recent items update automatically from the channel.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.