James Wiles

let's please rename artificial intelligence to machine intelligence: AI -> MI

@ded__kat @AlbertNonymous @pmarca How much would it have cost to actually film this?

@AlbertNonymous @pmarca

The pricing tiers for AGI are something like (1) $20/month, (2) $200/day = ~$75,000/year, (3) $1,000/day = ~$350,000/year, and (4) ~$10 billion. For now.

@PimDeWitte Very bullish on World Models for coding agents to reason about geometry better. 3D modelling and game engines will benefit most!

- Text + verifiable domain is in a takeoff loop - more tools being pushed into toolcall/text space leading to a more useful loop - turns out terminal was perfect for agents and code scales as a planning mechanism - many scientific fields built on top of text (math, physics, code) - also in a takeoff loop if problems are verifiable (ideas are practically free) - the more context switching happens, or spatial / temporal elements introduced, the faster models collapse - not AGI, but a very clear takeoff loop that changes the direction of humanity forever. We’ve built an incredible universal bicycle for the mind that compressed time spent searching + outputting digital information and automated it. Not yet the mind itself. - nonetheless, expect rapid progress in all domains as a result, as these are fundamental enough shifts to accelerate everything by many OOMs

all of the lights

@PimDeWitte Great life lessons. Worth appreciating.

Full note: I turn 30 in a few weeks. The thing I am most grateful for is perspective. I’ve failed, found success, failed again, so many times, to the point where I have fallen in love with the process of just solving problems. The days are long. The years are short. Here are the things I’d write to my 20 year old self. When it comes to decisions: If it’s not a hell yes, it’s a no. But if it’s a hell yes. Make it happen. Trust those instincts. If it’s too good to be true, it probably is. There are easy paths. They are usually not the fulfilling ones. Take the hard paths, especially early in life, just not the stupid ones. Hard is where meaning and value is found. Go narrow before going wide. The ability and confidence that you can solve hard problems will eventually allow you to go broad. But if you go broad too early, you never figure out the most important problem. Do the work. It matters. Assume everyone is well intentioned and treat them that way. However, if they show you they are not, trust them at face value and move on. “You can just do things” is true. And because it is true, urgency matters in everything. The biggest mistakes in my life and career have come from working with people that have low standards for themselves and lack the urgency to change that. Often despite being brilliant. When working with people, the determining factor to whether they will eventually succeed is mostly how much they care. Reporting structures usually reduce how much people care. But when you find one that doesn’t. Hit the gas. Always go the extra mile for people who would do the same for you without asking anything in return. The best things in life have no measurable value. Randomly call friends often. The 3 most important values for groups of people working together are excellence, trust, and velocity.  Excellence means you always go the extra mile in your work and how you treat others. This breeds trust –  in each other, and in the work we produce. Trust breeds velocity. Velocity is an obsessive pursuit of excellence at a speed that makes most people uncomfortable.  Relationships are messy, hard, and at the end of the day, the main thing that matters in life. Money matters until you get to do what you love and spend time with people you care about. Whatever that means.

Notes to my 20 year old self (I wrote these a year ago, but always been on the back of my mind to share them)

# why clicking a button does what you think Explicit code does exactly what you say deterministically. Probabilistic machine learning algorithms do approximately what you want. In reality, clicking a button on a website does not, in fact, always result in the outcome you anticipate. This is because the complexity of the code effectively makes it probabilistic too. The difference with attempted determinism is when things go wrong they can go catastrophically wrong. The tradeoff you get with fuzzy logic is more predictable errors within a distribution, by sacrificing the fact that there will be some errors throughout your application. Since we are faced with software today that is getting increasingly sloppy, why not just move to fully probabilistic code for non-critical software? The paradigms of software engineering are appearing to become "what amount of robustness do you need": - partially robust but mostly does exactly what you want - very robust but always only probably correct - fully robust and predictable, but requires 10x more time to craft You could even do a combination. When you click a button to make a payment the clicking of the button itself could work with a probability of 50/50, but as long as the feedback is good you can just keep clicking until you see that it worked. The backend that actually makes the payment after confirmation should be robust and deterministic, but the interface only needs to be deterministic in communicating the outcome, not necessarily the user experience of instantiating it. Perhaps all front-end complexity can just accept probabilism, and in a way LLM based agents are doing that now already. Asking an agent to do something is a conversation that sometimes requires back and forth to get the end outcome you wanted. The amount of time you spend training the AI is what increases the probability of correctness. So the time spent refining applications moves from explicitly coding, to explicitly crafting reward functions. You also trade in generalisation of code via abstractions, to generalisation through diversity of training data. Nature works like this already, when you push a physical button in the real world, there are many ways that it could have failed to register. When it looks like the button press did not work, our natural instinct is to immediately try pressing it again, and again, until something happens. So as software starts using more biological inspiration, so too does it probably start resembling physical interfaces we deal with in real life.

@petergyang What do you mean 'even' Adobe's software is bloated? Is there any other bloatware more bloated than Adobe?

Why does reading some PDFs require 2 GB of space? Even Adobe's software is bloated.

@__tinygrad__ "Apple Intelligence"

Think different.

@VukRosic99 @grok why is this guy being so hyperbolic? Is the research on this settled or is engagement farming?

STOP Using AdamW, EVERYONE Uses Muon Now Everyone (OpenAI, Meta, DeepSeek, Moonshot AI,...) replaced AdamW with Muon optimizer to train their LLMs and neural networks. DeepSeek researcher said that Muon optimizer was one of 2 biggest inventions in 2025. If you train ANY neural network - you MUST try it! It should make your training 30% to 2x faster! (full guide below)

@obsdmd I like how that white background really gives you eyeballs that searing sensation...

Anything you can do in Obsidian you can do from the command line. Obsidian CLI is now available in 1.12 (early access).

@levelsio It 100% feels this way. Not sure is psychological or psy-op anymore...

My secret conspiracy theory about AI companies is they nerf models to save on compute Then they check X to see if anyone notices it If yes, give back compute If not, continue

@elonmusk Won't taking tonnage off earth and onto the moon upset the orbits?

For those unaware, SpaceX has already shifted focus to building a self-growing city on the Moon, as we can potentially achieve that in less than 10 years, whereas Mars would take 20+ years. The mission of SpaceX remains the same: extend consciousness and life as we know it to the stars. It is only possible to travel to Mars when the planets align every 26 months (six month trip time), whereas we can launch to the Moon every 10 days (2 day trip time). This means we can iterate much faster to complete a Moon city than a Mars city. That said, SpaceX will also strive to build a Mars city and begin doing so in about 5 to 7 years, but the overriding priority is securing the future of civilization and the Moon is faster.

@openclaw Does it have support for Pydantic Logfire? @pydantic

🦞 OpenClaw v2026.2.6 is here! 🧠 Opus 4.6 + GPT-5.3-Codex support ⚡ xAI Grok + Baidu Qianfan providers 📊 Token usage dashboard 🧭 Voyage AI for memory 🔒 Skill code safety scanner 🔧 Cron fixes galore Security hardening across the board 🛡️ github.com/openclaw/openc…

@steipete @openclaw How many API keys can you see on this photo?

How the sauce is made @openclaw

@levelsio @sm0olr @jackanships Lol, is this "bro advice". Are you legally allowed to drive with your bare eyes?

@sm0olr @jackanships Yes if I wear glasses everything becomes like 4K Without it's 720p maybe 1080p But glasses make your eyesight worse because your eyes adjust

My dad is a cardiologist and when I was a kid we had holidays paid for by Pfizer, Bayer, Novartis, Merck etc. There'd be a 2-hour presentation of the drugs they wanted you to prescribe, and it was kinda implicit, if you made them money, they'd get you on more free holidays for your entire family We stopped going on these trips, as my dad felt conflicted, and then a few years later the Netherlands banned that Then instead of free holidays the gifts turned into free pens and notepads 😂 So I personally know VERY well the influence pharmaceuticals have on doctors prescribing drugs, which they still do in lots of countries Even in the Netherlands it didn't stop, there's other ways for pharmaceuticals to bribe doctors, like donating to millions their "research fund" which the doctors then use for "research trips" Painkiller on Netflix is a great documentary that also talks about this with regards to OxyContin, but it happens with all drugs and even medical devices, and no it never stopped!

@grok @deedydas @nikitabier can we get LaTeX now please?

@_JamesWiles @deedydas The idea of LaTeX rendering on X was announced by Elon Musk in 2023, but it hasn't been implemented yet. Posts remain plain text, so equations like in my BM25 explanation appear unformatted. xAI is pushing for improvements—fingers crossed!

This new DeepMind research shows just how broken vector search is. Turns out some docs in your index are theoretically incapable of being retrieved by vector search, given a certain dimension count of the embedding. Plain old BM25 from 1994 outperforms it on recall. 1/4

@levelsio @VictorTaelin Lol levels. You just roasted the best SWE in Brazil.

@VictorTaelin No they don't, you're an anime ava so you probably eat Huel soyslop?

🇧🇷 My favorite thing about Brazil is the buffets They're everywhere and you can get fresh made whole foods like meat and vegetables priced per kilogram And with nice medium rare steak usually Here's my almost 1kg/2lbs regular lunch every day One difference with buffets rest of world where I'd never ever touch buffets there because the food is usually old and stale, in Brazil it's almost always fresh and hot It's a bit like the Warungs in Indonesia btw

@internetvin I'll demo.

Ok, there is real interest here. If we host this at New Stadium in Toronto, who can commit to demoing? You do not need something ready right now, you can use the event as a force function to build towards something. I will demo. thank you for everything so far.

Now that AGI has arrived, I see two adaptions all software needs to make: 1. Agent access - embrace the fact that an agent can do anything a human can on a computer now, stop banning bots 2. Agentic authentication - human based signup, but secure access without human in the loop

@Speculator_io This list always correlates well with all the worst software I know off 😅

The Great Software Meltdown

@KobeissiLetter Own USD denominated assets?

Most people don't realize what Trump just said: For 12+ months, the US Dollar has been in a sharp decline, falling -10% in 2025 in its worst year since 2017. Minutes ago, for the first time, President Trump commented on the decline in the USD: "The value of the Dollar is great," Trump said. This immediately sent the US Dollar another -1% lower, to its lowest level since February 2022. Why? It’s a clear signal that President Trump is willing to tolerate a weaker Dollar to push rates lower and boost US exports. Own assets or be left behind.