Gray Newfield retweeted
Own your intelligence.
Anyone that opposes this has something to sell you (perhaps your own intelligence back to you.)
Local models need to continue proliferating. The “absolute unit” massive models may be out of reach for now locally, so your best option is to, indeed, keep your options open. Own your harness, data, connections, runtime: your agents. Embrace open source.
Anyone that opposes this has something to sell you (perhaps your own intelligence back to you.)
Local models need to continue proliferating. The “absolute unit” massive models may be out of reach for now locally, so your best option is to, indeed, keep your options open. Own your harness, data, connections, runtime: your agents. Embrace open source.
194
186
1,784
81,414
Gray Newfield retweeted
Beautiful
A ‘hello world’ milestone for Project Suncatcher 🚀
After launching our prototype satellite a few weeks ago, we’ve now successfully sent a command up to orbit, booted up the TPU, ran the prompt, and received our first complete response back on earth - a haiku!
0
1
1
235
Gray Newfield retweeted
My further (informed) fantasies about the near future:
1) No CI. Agents prove each change against rules while they write it.
2) Code is a cache: rules stay, code regenerated with better models.
3) No staging, simulation instead.
4) Humans direct in plain words, agents compile that into formal rules, humans don't read the formulas, only examples and counterexamples.
5) Agents find the rules the human never knew to express (and ask to confirm them).
6) The checker catches contradictions between human's rules while they're still directing.
7) Unchecked/unproved code becomes legacy code.
8) Branches come back as experiments (many versions at once, humans pick among winners).
9) Software gets bred, not built.
10) Slop is variation without selection.
11) If UI is generated on the fly, design becomes the rules every generated screen must follow. A checker runs those rules on every generation.
12) Apps evolve from how they're used.
13) Everyone uses their own software version. One product becomes many personal versions, that stay compatible via shared rules.
1) No CI. Agents prove each change against rules while they write it.
2) Code is a cache: rules stay, code regenerated with better models.
3) No staging, simulation instead.
4) Humans direct in plain words, agents compile that into formal rules, humans don't read the formulas, only examples and counterexamples.
5) Agents find the rules the human never knew to express (and ask to confirm them).
6) The checker catches contradictions between human's rules while they're still directing.
7) Unchecked/unproved code becomes legacy code.
8) Branches come back as experiments (many versions at once, humans pick among winners).
9) Software gets bred, not built.
10) Slop is variation without selection.
11) If UI is generated on the fly, design becomes the rules every generated screen must follow. A checker runs those rules on every generation.
12) Apps evolve from how they're used.
13) Everyone uses their own software version. One product becomes many personal versions, that stay compatible via shared rules.
Next question: why are you still using branches?
We talked about PRs. I don't think they make sense anymore, if you rethink them from first principles.
Yes, if you only use them to trigger CI, then sure, go ahead. That's a fat commit, not a PR. But for mandatory code review? Knowledge sharing? C'mon.
But some of you said: "we use PRs because we use branches! I love to keep my stuff on branches."
And I'm here to tell you: you don't need branches! You can just push to main.
It's called trunk-based development and the idea behind it is true continuous integration: you continuously integrate your changes with the trunk, the main branch.
It's been around for a while, it works. Just yesterday I talked with someone who said they've used it in an org with 400 engineers, without any issues. Ruby, the project, uses trunk-based development.
The main idea is that you want to avoid long-lived branches. They're bad, because branches can diverge and merging can turn into a thing. And you don't want routine operations to become a thing.
Even when I worked in teams that used PRs and branches, I considered it a smell when a branch lived for longer than a day or two.
But when I shared this with others, I'd often get something back like this:
"Well, good for you, but no. MY thing, what I'm working on, that can't be done on main. It needs to be on a branch! It has a database migration, it changes the underlying shape, the architecture, it refactors this and changes that. It can't go on main yet! I need to merge it all at the end."
And then I'd say:
"Nah, I can merge nearly anything continuously in small commits to main. You split things up. Add the column first, merge that, deploy it, now it's out, works, safe.
Add the helpers first, merge them, get them out.
Want to migrate from legacy pattern A to new pattern B? Don't do it all at once. Add the dependency, then migrate A to B, one by one, safe."
If it can be merged in a large change, I bet it can be done in small increments that you merge to main.
Yes, there are exceptions, but it's 0.5% of cases, not 50%.
But there was a real problem with this approach: it was hard!
There's a lot of "meta programming" you have to do: make sure this small first change compiles and runs and the tests work, merge it, get it out, monitor it; then do the next small piece, make sure that works and get it out; then maybe push some still-unused dead code, add a feature flag, add a config variable; or keep both configurations working at the same time, etc.
That meant sometimes it was just easier to do it in a big fat branch. But big fat branches are risky!
A system that continuously flows is safer than a system that starts and stops in spurts.
But -- you guessed it -- here's the good news:
Now, with agents, all of these chores are no chores at all!
Agents are great at adding feature flags, they're great at splitting up code, they're great at keeping track of three changes and merging them one by one and watching them roll out. They're great at keeping two configurations working for a day. They're great at migrating one by one.
In other words: it's now 10x easier to just work on main and make sure everything is continuously integrated and working.
It's now easy to make sure the system constantly flows.
And look, ...
I'm not here to convince you to switch to trunk-based development. Just a ruse to get you to read this. (But also, hey: try it! Pushing to main is amazing.)
I'm here to make you RETHINK how you develop software.
Don't just use agents to replace your typing.
Use agents as the magic code-writing machines they are. Everything that was hard about software development because it involved writing code -- even the meta programming -- has become easier.
You can now reach for the stars! But you need to look up.
32
23
397
30,072
Started using @t3dotcodes this week, and it's awesome. Proof that, in the age of AI, two things are true:
1. A small team can make a great product.
2. The only way to make a great product is to build for yourself, talk to customers, and ship fast.
Good job @theo @jullerino!
1. A small team can make a great product.
2. The only way to make a great product is to build for yourself, talk to customers, and ship fast.
Good job @theo @jullerino!
1
0
4
259
YC is the YC for crypto
At @ycombinator, we're excited to fund more crypto startups solving financial needs including AI capital raising, stablecoins, trading, and agents.
Here's our latest crypto request for startups. Our next application deadline is Monday.
0
0
6
349
100% agree. Last week, I promised myself I would stop pouring time into my pc / network setup – and I kept it for a couple days – but then I dove back in again. It’s more than curiosity. I think I'm drawn to this problem because I feel like we’ve been patiently waiting for the models to get good enough to do anything. They’ve been smart for a long time, but there have always been things they could do better. However with Opus 5.5 and GPT 6.1, I think we’ve reached an inflection point. The bottleneck for me is no longer the model or harness but everything else around it: the hardware and OS it’s running on, the system settings, the browser it uses, and the access pattern for passwords and 2FA. The most crucial is the interface for interacting with it, but the reason has flipped: previously, I wanted to optimize for my ability to steer the agents – now it’s about the agents’ ability to steer me. The thing they can’t do without me is make decisions, so the thing I need now more than anything else is a way for the agent to help me make the best decision I can as quickly as possible. I’m the critical section of the system now, and there needs to be a mutex lock on accessing me because I can only make one decision at a time. Therefore the most important thing is giving the agent a way to load my context with all the info I need to make a decision, provide a way to reason about it (ie diagram, mock, prototype, freeform, etc.), and let me convey that decision back to the agent with as little or as much nuance as I want. I’ve actually found that ChatGPT’s IntelligentUI and Claude Code’s Artifacts are pretty good at this, so I’ve gotten in the habit of treating a single Chat or Claude thread as my “coordinator” and having it mediate between me and the worker threads, since it understands how to communicate effectively with both of us. The urgency I feel for solving this is so high because everyday I don’t have the right setup is a day of missing out on the value these models can create, while someone else out there is getting it and the gap between us is growing. The question everyone jokingly asks is, “Is this AGI?” I think the answer is it depends on who’s using it.
FYI here are the tools I'm currently using that have gotten me closest to the ideal setup. Highly recommend using them:
– @1Password
– @AsideAI (browser w/ MCP for any agent)
– @t3dotcodes (thread management)
– @bot (coordination and project management)
Also, a huge unlock for me was moving my development to a Linux machine (I installed it on an old T2 Macbook Pro). You don't realize how much MacOS is handicapping the agents rn. I also generally avoid Claude for anything that's not strictly engineering because it insists on permission prompts 4x more frequently than every other agent, and it's completely unnecessary.
TLDR; Most people don't realize this, but under the hood, we're all driving Ferrari's. The problem is that, even if you figure this out, the next thing you realize is that you've been driving through a school zone for the last 250 miles. You need to grab the wheel and get on the Autobahn as quickly as you can.
FYI here are the tools I'm currently using that have gotten me closest to the ideal setup. Highly recommend using them:
– @1Password
– @AsideAI (browser w/ MCP for any agent)
– @t3dotcodes (thread management)
– @bot (coordination and project management)
Also, a huge unlock for me was moving my development to a Linux machine (I installed it on an old T2 Macbook Pro). You don't realize how much MacOS is handicapping the agents rn. I also generally avoid Claude for anything that's not strictly engineering because it insists on permission prompts 4x more frequently than every other agent, and it's completely unnecessary.
TLDR; Most people don't realize this, but under the hood, we're all driving Ferrari's. The problem is that, even if you figure this out, the next thing you realize is that you've been driving through a school zone for the last 250 miles. You need to grab the wheel and get on the Autobahn as quickly as you can.
I no longer get frustrated that LLMs feel dumb. Until Fable and Opus 5.5, some tasks were beyond LLMs no matter how hard I pushed. Not anymore. These two models feel smart enough for everything I want them to do.
What slows me down now:
- Not knowing my context and preferences
- Permission prompts and refusals
- Passwords and 2FA in the browser
- Inexperience (no RL data for some domains)
- Claude speak and UI slop
All but the last seem solvable with good engineering, no further AI breakthroughs needed.
1
0
5
383
Gray Newfield retweeted
I no longer get frustrated that LLMs feel dumb. Until Fable and Opus 5.5, some tasks were beyond LLMs no matter how hard I pushed. Not anymore. These two models feel smart enough for everything I want them to do.
What slows me down now:
- Not knowing my context and preferences
- Permission prompts and refusals
- Passwords and 2FA in the browser
- Inexperience (no RL data for some domains)
- Claude speak and UI slop
All but the last seem solvable with good engineering, no further AI breakthroughs needed.
What slows me down now:
- Not knowing my context and preferences
- Permission prompts and refusals
- Passwords and 2FA in the browser
- Inexperience (no RL data for some domains)
- Claude speak and UI slop
All but the last seem solvable with good engineering, no further AI breakthroughs needed.
16
7
75
7,687
The UX I want:
– One personified agent
– One chat
– Reply to either main chat or inline thread
I do not need or want:
– A team of named agents
– Tabs, side chats, or channels
Claude Projects is the closest rn. Eager to try Duet, but could be one step forward, two back.
– One personified agent
– One chat
– Reply to either main chat or inline thread
I do not need or want:
– A team of named agents
– Tabs, side chats, or channels
Claude Projects is the closest rn. Eager to try Duet, but could be one step forward, two back.
0
0
4
205
The US is just a country. America – the West – is an idea. It's the belief that individual freedom is the only thing you need in order to build a better life for yourself and your family.
Whether it was pilgrims crossing the ocean or settlers following the Oregon Trail, this simple but powerful idea has inspired countless generations to leave everything behind in search of a new world. Millions of people from all over the map continue to risk everything in pursuit of this dream.
We are no different. We've just forgotten. We are still pioneers. We will seek out the frontier wherever it may be. Whether it's on the moon or in the metaverse, our manifest destiny is waiting there for us. We must fulfill it, not for our own sake, but for everyone who comes after us.
The future isn't a time or a place. It's a People. They don't get a vote, but they do have a voice. If you can hear it, you'll become an optimist. If you listen to it, you'll become a builder. If you follow it, you'll become a founder.
That's all there is to do. It's more than enough for me.
Whether it was pilgrims crossing the ocean or settlers following the Oregon Trail, this simple but powerful idea has inspired countless generations to leave everything behind in search of a new world. Millions of people from all over the map continue to risk everything in pursuit of this dream.
We are no different. We've just forgotten. We are still pioneers. We will seek out the frontier wherever it may be. Whether it's on the moon or in the metaverse, our manifest destiny is waiting there for us. We must fulfill it, not for our own sake, but for everyone who comes after us.
The future isn't a time or a place. It's a People. They don't get a vote, but they do have a voice. If you can hear it, you'll become an optimist. If you listen to it, you'll become a builder. If you follow it, you'll become a founder.
That's all there is to do. It's more than enough for me.
Thiel: “There is no successor. When the US is done, the West is done, and the world is done.” — if you truly believed this, what would you do?
0
0
1
127
Is it just me or has Amazon really gone down the tubes? Higher prices, unreliable reviews, junkier products, defective items, slower shipping promises, late deliveries, unhelpful customer service... Sad.
0
0
4
140
Gray Newfield retweeted
First they ignore you, then they laugh at you, then they fight at you, then they add you to their Long-Term Capital Markets Assumptions paper and predict you'll be the best-performing asset of the next decade.
192
933
5,324
462,235
💯
Something that’s often overlooked on the ROI of spending more via LLM tokens to generate bespoke replacements of Saas (for internal use) is that Saas by definition is the lowest common denominator. Very few offer the ability to customize. Claude generated solutions by definition include the features I personally need. Maybe it’s not perfect or production-grade, but it does the job better and I can iterate with new features as needed. This produces a far higher ROI than working around the limitations of a $25/mo software solution.
For example, I needed to draw and prompt edits on the canvas and Claude added that within 2 minutes. No feature request, no support chat. Just one prompt and it’s live.
0
0
1
191
I've been having a blast vibecoding little tools for myself
It's productive, fun, and it honestly helps recalibrate your swing as a founder: build→talk to customers is easy when you're the customer!
If you want to make something people want, start by making something you want
It's productive, fun, and it honestly helps recalibrate your swing as a founder: build→talk to customers is easy when you're the customer!
If you want to make something people want, start by making something you want
Inspired by the work @benjaminsehl's team has done with the new Shopify Canvas editor, I had Opus 5.5 build an internal tool for viewing screenshots of every state of my iOS app. Quickly switch between light and dark mode as well go back in time to previous states.
It's a nice ol. This is what I love about Opus 5.5 and the state of software development. It's basically free and insanely quick to whip up small tools to improve common workflows.
1
0
4
335
There are only a handful of places in the world that have some immunity to this kind of chaos, and I don't think the West can successfully emulate any of them:
1. China – Strong authoritarian government with no individual rights
2. Japan – Conservative democracy with extremely tight migration controls
3. Israel – Strong sense of national identity and civic duty cultivated through mandatory public service and ancient cultural heritage
4. Singapore – Classical liberal values, extremely competent bureaucracy, and Draconian law and order policies that maintain stability
5. Dubai – Highly stratified society with ethnically homogenous and wealthy class of elites, insulated by strict theocratic / dictatorial government
Western Europe has already welcomed mass migration, so they can't be Japan. Rampant antisemitism plus white guilt means they lack the conviction, historical memory, or love of country to be Israel. They don't have the competence or technological prowess to be Singapore. And they lack the wealth or natural resources to be Dubai.
That leaves only one option. They can't be China, but they can become clients of China.
Russia continues to threaten them from the East. And out of pride or arrogance, they've rejected their US ally in the West. So the only hegemon they can realistically align themselves with in this tri-polar world is China, which they already depend on for trade and investment.
My prediction is that the governments of France, UK, Germany, etc. will try to subdue internal unrest as long as they can, until they're eventually transitioned out of power by a wave of populism. That new, radical government (can be either right or left – it doesn't matter) will fail spectacularly to solve any of the economic, social, or political problems they campaigned on. Their economies will bottom out, and they'll have only three sovereigns to turn to. China will be the most eager suitor, if not only to seize a few technological gems, like ASML, stick it to the US, and throw off the global balance of power.
IMO this has been China's plan all along and why they've supported Russia's invasion of Ukraine, Palestinian terrorism against Israel, and Iran's war against the US. Each one of these issues has further polarized the US at home and driven a wedge between us and our NATO allies abroad.
TLDR; We are undeniably watching the Fourth Turning unfold. Our goal should not be to save these institutions (futile). Instead, our best course of action is to race towards the singularity and use artificial intelligence and blockchain technology to build up a decentralized network state, rededicated to classical liberalism, freedom, and democracy.
I may be completely wrong, but I'd still choose to spend my days the same way, building the technology that has the highest probability of unlocking a prosperous future for human civilization. In every era, there should be people who aim for the stars, even if it seems like we'll fall short. The cynics will call us fools and ask why it matters how we fall. But when the fall is all that's left, it matters very much.
1. China – Strong authoritarian government with no individual rights
2. Japan – Conservative democracy with extremely tight migration controls
3. Israel – Strong sense of national identity and civic duty cultivated through mandatory public service and ancient cultural heritage
4. Singapore – Classical liberal values, extremely competent bureaucracy, and Draconian law and order policies that maintain stability
5. Dubai – Highly stratified society with ethnically homogenous and wealthy class of elites, insulated by strict theocratic / dictatorial government
Western Europe has already welcomed mass migration, so they can't be Japan. Rampant antisemitism plus white guilt means they lack the conviction, historical memory, or love of country to be Israel. They don't have the competence or technological prowess to be Singapore. And they lack the wealth or natural resources to be Dubai.
That leaves only one option. They can't be China, but they can become clients of China.
Russia continues to threaten them from the East. And out of pride or arrogance, they've rejected their US ally in the West. So the only hegemon they can realistically align themselves with in this tri-polar world is China, which they already depend on for trade and investment.
My prediction is that the governments of France, UK, Germany, etc. will try to subdue internal unrest as long as they can, until they're eventually transitioned out of power by a wave of populism. That new, radical government (can be either right or left – it doesn't matter) will fail spectacularly to solve any of the economic, social, or political problems they campaigned on. Their economies will bottom out, and they'll have only three sovereigns to turn to. China will be the most eager suitor, if not only to seize a few technological gems, like ASML, stick it to the US, and throw off the global balance of power.
IMO this has been China's plan all along and why they've supported Russia's invasion of Ukraine, Palestinian terrorism against Israel, and Iran's war against the US. Each one of these issues has further polarized the US at home and driven a wedge between us and our NATO allies abroad.
TLDR; We are undeniably watching the Fourth Turning unfold. Our goal should not be to save these institutions (futile). Instead, our best course of action is to race towards the singularity and use artificial intelligence and blockchain technology to build up a decentralized network state, rededicated to classical liberalism, freedom, and democracy.
I may be completely wrong, but I'd still choose to spend my days the same way, building the technology that has the highest probability of unlocking a prosperous future for human civilization. In every era, there should be people who aim for the stars, even if it seems like we'll fall short. The cynics will call us fools and ask why it matters how we fall. But when the fall is all that's left, it matters very much.
Perhaps the best thing to remember when we see riots and arson attacks on historic buildings in France, or hundreds of drug addicts sprawled out on the streets of San Francisco, or thousands of murders per year in America’s cities, is that things don’t have to be this way. There are places in the world that you can visit that are not like this. There are a set of policy choices that we can make at any time that would ensure these things simply don’t happen, if our civilization had the will to enact them
1
0
4
289
Gray Newfield retweeted
We're going to look back at one-size-fits-all software the way we look at the three-channel TV era
59
60
1,334
48,828
The perfect bathroom doesn't exi–
In honor of the Chrysler Bldg trading I’d like to share that my incessant tweeting asking for a tour actually worked.
A stranger (now friend) on ReTwit hooked it up: Mr. Chrysler's private bathroom (69th floor), the 67th-floor “Dentist in The Sky”, the 🦅 s , and a ladder climb up to the tippyyy top.
0
0
0
132