Joined December 2014
🚹 CHINA JUST OPEN SOURCED AN OCR THAT EATS 100 PAGE PDFS IN ONE PASS It’s called Unlimited OCR. Only 3B params. Runs locally. Every other OCR tool chops your doc into pages and loses the thread. This one reads the whole thing in one continuous pass. → One shot full context parsing (32K context window) → Multilingual out of the box → 93% on the standard parsing benchmark (6 points over baseline) → Under 0.11 error rate past 40 pages → Runs 100% locally on your own hardware → Works with Transformers, vLLM, SGLang, Docker, Ollama, llama.cpp Traditional cloud OCR (Textract, Google Vision, Azure Doc Intelligence) costs $1.50 to $15 per 1,000 pages. This runs on your machine. For free. Forever. Baidu built it explicitly to push DeepSeek OCR one step further. Already at 1.9M downloads on Hugging Face and most people have no idea it exists yet. 100% open source.
144
986
50
9,351
492,094
Clumsy Pilot retweeted
Replying to @jasonc_nc
In the same vein:
1
1
47
911
Everybody’s missing Tao’s whole point. Imagine an AI-generated cancer drug that’s specifically designed to game the drug trial process. The drug makes the cancer cells go dormant and changes the way they appear in blood tests and diagnostic imaging, such that they become indistinguishable from healthy tissue. The drug clears the trial and then only much later do we find out it didn’t work. It probably was also harmful on top of that. We *already* see AI behave like that when confronted with difficult programming/hacking challenges. This is already reality.
This is a slide from Terence Tao's 'Math 2.0' public lecture yesterday evening at Caltech. I am highly confident that, if the cure actually worked, almost any Stage 4 cancer patient would tell you that human comprehension of its mechanism is completely irrelevant to them.
369
779
107
7,836
242,497
Clumsy Pilot retweeted
A majority of Walt Disney World traffic is childless adults Catering to adults increased revenues and profits as childless adults have more disposable income to spend on themselves. Thus, infantile adults have priced out many actual children, who make up a dwindling minority of the park's guests and who are by virtue of astronomical prices, restricted by class. The same dynamic has already transformed the toy industry: adult-only households account for 55 percent of spending on toys.
221
503
111
5,280
1,229,019
Clumsy Pilot retweeted
In real terms, inflated by wages, Brunel’s Clifton Suspension Bridge cost about £116m in today’s money, about half of the cost of repairing Botley Road bridge in Oxford. What is going on ? createstreets.com/the-botley

ÂŁ237 million and they couldn't even get matching railings. Another example of how, even in the most gold-plated expensive projects, beauty is not part of the equation, despite the massive impact it has on our wellbeing.
17
47
6
413
19,880
Clumsy Pilot retweeted
Grim reading from Chris Gibb on HS2. It’s hard to comprehend the £100bn spend and yet be faced with this potential outcome - slower journeys to Glasgow, fewer seats, no tilt, no kitchens, no room for growth. Do the public comprehend what a total disaster this project has been?
22
57
14
329
33,414
Clumsy Pilot retweeted
Shelagh Fogarty arguing with two Jewish women on LBC, one of whom is a holocaust survivor. Shelagh’s unwavering support for the Netanyahu government is quite unbelievable. To interrupt and cut off these two women because they dared to be critical of Israel is mind boggling.
133
516
12
2,787
48,712
me when my vitamin d was 10 and i kept saying “my bones hurt.. everything hurts..” and everyone around me said “that’s not possible” but my doctor immediately asked “don’t your bones hurt? aren’t you in some kind of pain?” and everyone had to apologize to me for months
found out i'm deficient in vitamin b12 and this is genuinely how it feels
86
1,905
198
48,290
896,027
the real reason nobody can build an american shenzhen is that western capital fundamentally despises low-margin physical infrastructure we spent 30 years financializing our economy & replacing hardware engineers with software middle management so now we have infinite capital for app wrappers but 0 physical muscle memory left to assemble the world you cant legislate industrial density when your entire elite class is trained to extract yield instead of moving atoms
china has already won the hardware war & most of the west just hasn’t processed the lag yet the whole gap is architectural shenzhen built a distributed neural network of manufacturing intelligence where thousands of specialized nodes iterate in parallel compounding knowledge horizontally every single day while a founder in europe spends 12k & waits two months for a single board revision you burn through four iterations in a week in huaqiangbei for a fraction of the cost the game is already over the fundamental disconnect comes from viewing manufacturing as a back office cost center to optimize & outsource while China treats it as a high density real time learning organism when your supplier modifies your injection mold tooling before you even open your mouth because his local network has already seen a hundred founders face the exact same thermal bottleneck the intelligence no longer lives inside a single company it lives in the ecosystem itself trying to out compete that level of information density with legacy corporate processes is like telling a horse to gallop faster when the other side just invented the combustion engine we are watching the most asymmetric economic competition in modern history & western leaders are still trying to solve an ontological crisis with management level tweaks tier 2 cities like hefei & Wuhan are already replicating this hyper dense hardware model at scale scaling ev & semiconductor infrastructure while europe debates its next regulatory committee you cannot patch a broken architecture with longer hours until the west stops optimizing variables that china stopped tracking years ago it will keep arriving at the wrong destination just slightly faster
67
506
40
3,763
411,302
Clumsy Pilot retweeted
Palantir just released a new surveillance system that fuses satellites, radars and every camera into one real-time tracker of people and movement We have to consider the possibility (probability?) that with Flock as the eyes, a system like this could be overlaid onto the US and we could be the targets. Another reason we cannot allow Flock to exist.
563
4,703
696
13,061
2,316,424
Clumsy Pilot retweeted
The anti-datacenter movement is literally the most rational response imaginable to an industry whose leaders have publicly said the best case scenario for their product is to get everyone fired. If you can’t see this then your brain is in late stage tech rot.
The anti-datacenter psychosis sweeping through both sides of our national politics might be the worst mass movement I've seen in my lifetime. The magnitude of lies underpinning it, the threat it poses to our economy and civilization, and the sheer scale of its popularity are totally unprecedented. It should be an enduring national shame that the country that invented the airplane, landed on the Moon, and created the Internet is descending into superstitious panic at the very moment growth matters most
416
1,451
153
14,711
809,130
To me it is peculiar that some of these deep thinkers: -find it totally normal and logical for the US to tell ASML and TSMC to not sell advanced stuff to China ("or else..") -then find outrageous and villainous and against the Rules Based World Order the mere thought of China ever telling TSMC to not sell advanced stuff to US "or else.." I can already hear it "nnoo u dont undertanddd, totally different stufff". Sure sure, man, playing half court tennis? btw i find BOTH cases totally logical and totally outrageous at the same time😂
Ben Thompson: "I think it would be very problematic for the US to win the AI race." "Take the scenario, where if you control AI, your military is better than anyone else's. What is the game theory optimal response of China? To blow up TSMC. There's a fundamental disconnect that I have with a lot of the rhetoric coming out of Silicon Valley. If we get to a place where we have a meaningful superiority, particularly from a military national security perspective, I think that's very dangerous for the world. The only world I see where we truly pull out and have no dependency on China, such that if they want to take Taiwan, who cares, seems pretty fantastical to me. There's a bit of facing reality in this regard that is not present in these conversations."
2
6
58
13,642
Wow. Proponents of the ICE taser gloves are openly saying that one reason they love the gloves is that they doesn’t leave any marks or proof that they were used, making it harder for people to sue when they’re shocked excessively.
201
6,469
185
16,342
463,817
Clumsy Pilot retweeted
The yacht of Mark Zuckerberg allegedly refused repeated calls by US Coast Guard looking for ships available to rescue a small boat with 2 women, a kid and a dog. According to a user on BSKY who was on a cruise ship that performed the rescue eventually. AIS-track fits claim.
362
6,328
324
28,872
1,303,144
Poor on purpose
New Syn Post: Embodied Insanity Bricks are cheap, aesthetic and still largely produced domestically. But new standards and dubious carbon arguments are pushing us to timber, 80% of which we import.
1
84
Clumsy Pilot retweeted
Why are we so ALLERGIC to change? This leisure centre is decaying. The council wants a new one. The new building will have a larger swimming pool, a café that overlooks the seafront, and a soft play area. Naturally 534 people objected. Why do we treat improvement as a CRIME?
5
9
140
5,282
One uniquely exhausting thing about writing on China is that every proper argument seems to require preparing a mini Wikipedia of caveats, footnotes, and pre-emptive clarifications. You can't simply make the argument you actually want to make. You have to first ask yourself what assumptions your readers will bring to the table or what common misconceptions or debates might be triggered by a particular phrasing. Trying to make a good faith effort of it requires trying to anticipate as many possible objections as possible before they occur and getting ahead of them. As an example from my piece today, I wrote that industrialization is not a queue, and that productive capacity does not automatically relocate to whichever country has the lowest wages once another country moves up the value chain - it also needs a foundation of productive capacity to receive it. To me, that seemed like a fairly straightforward point about development economics. One critical response I received: "China had nothing when we decided to industrialize them 50 years ago." So then we aren't discussing my argument anymore. We are now debating whether China itself possessed any meaningful industrial capacity prior to the reform era...something I had neglected to pre-empt in my piece. Suddenly I find myself wondering whether I should have included an extra paragraph in the price or at least. footnote explaining that China in 1979 was poor but not a barren wasteland. It already possessed infrastructure, industrial enterprises, power generation, engineers, railways, ports, literate human capital, state capacity etc. It was in fact unusually well positioned to receive FDI. I know this. China scholars know this. People that study development economics know this. But most people probably don't know this...so is it the critic's fault for being ignorant, or my fault for not foreseeing the issue? Of course even if I got that one, I'd probably miss another. The piece is still going. Maybe someone will object to how I've defined infrastructure or challenge my definition of industrialization itself. Or more likely an objection I can't even imagine yet. No matter how many caveats and qualifications you add, someone will always find a background assumption you neglected to footnote. The dispiriting thing here isn't that people will disagree with me. That's part of sharing an opinion in a public space. But it's a bummer that China discussions feels like they require reconstructing an entire agreed worldview from first principles before you can have any kind of conversation. I think that's probably something most people writing about China have experienced. Or at least will recognize.
91
160
31
1,312
109,383
Clumsy Pilot retweeted
We could’ve automatically dealt with a huge chunk of the triple lock ‘problem’ by simply allowing the State Pension to be taxed at 20% once it rose above the personal allowance, as it will this year. Instead, of course, Andy Burnham is going to introduce Sunak’s Quadruple Lock.
33
72
8
1,158
51,510
Clumsy Pilot retweeted
IMO reading CEO of Deepseek's interview and comparing it to what Western AI leaders are saying is the largest blackpill I've taken today
110
476
78
4,888
484,538
Clumsy Pilot retweeted
This blew up. And I think it's fair to say it touched a nerve. The comments I've had are mostly split between two answers
 either wipe out the shareholders, or nationalise the whole thing. Or both. And the belief is that the problem is then solved. It isn't. Here's why
 A quick point before I jump in. This looks like a Thames Water story, but in reality, Thames is just the first to hit the wall where everyone can see it. What follows is about all of it. First, I think we need to be clear on how this whole thing works. Ofwat has no money. All Ofwat does is decide how much the water company is allowed to charge you and exert influence over how the money is spent. So when we talk about investment in the industry, what we're really talking about is the water bills that are paid by you. There's no other pot of money here that can be dipped into other than what the government may provide in subsidies. Every 5 years, Ofwat sets the bills and then the companies propose what they'll do with the money. In the last cycle, the companies were allowed to increase your bills and charge you for replacing 0.4% of the water mains per year, which would mean our national water infrastructure would be fully replaced over 250 years. However, they only managed to replace at a rate of 0.14% per year, which is a full replacement period of 700 years. Or to put it another way, a period that takes us back to a time before the printing press existed. Those figures come from the NAO. So the companies increased your bills to pay for replacement water infrastructure, that they only delivered about a third of. Now, the hard-money, free-market side of me wants to say this is what price-fixing always does. When you set a central cap on what a company can charge, you distort every decision it makes. But that isn't the system we have. We don't have a free market in water, because you can't. Nobody's laying a rival set of pipes to your house, so you're stuck with the one option you have in front of you. What we've built instead is the worst of both worlds
 a private monopoly with all the profit motive of a business, sitting behind a price cap which has none of the discipline of competition. So the incentive isn't to serve the customer, because the customer can't leave, they're captive. And it isn't to invest for the long term, because like our electoral cycle, the cap resets every five years. So the incentive is to spend as little as you can get away with on the things nobody can see, and take the difference out now before inflation eats it away. Which is the story here. That's the design of the system. Now this is where it gets serious, and is why this bill is coming your way, one way or another. The NAO says £290bn is needed in investment to improve our water infrastructure by 2050, just to hit the government's targets. That's on top of day to day running costs, which are spiralling. Ofwat estimates another £52bn is needed for reservoirs and major supply projects. The government accepted the £290bn figure in its White Paper back in February, so these figures aren't in any sort of dispute. And this is where the £104bn figure you might have seen quoted comes in. That's the number the industry keeps waving around as proof they're finally investing in the infrastructure that we so badly need replaced. This is the five-year settlement Ofwat signed off in 2024 to start delivering new reservoirs, less leakage, and fewer sewage spills. This is why your bill is rising 36% this decade. You're paying for it, because there is no one else. But when we set it against that £290bn, it's barely a third of what's needed by 2050. So this isn't the whole fix. In fact, it's just a first instalment. And it only exists because your bills went up to cover it. So this is where you'll feel the sting, and it's a nasty one. You've already paid for the pipes once, and you got a third of what you were charged for. Now the investment needed is four times what it was, and you'll be told to be grateful for it when you receive the bill. So back to where we started
 wipe out the shareholders, their stake is worthless anyway. Give the bondholders a haircut. Nationalise the whole thing. Do all of it. Now, before anyone tells me ownership is the answer, take a look at Wales. Welsh Water has had no shareholders since 2001. To be precise, it's not nationalised
 it's a private not-for-profit, so there are no dividends, no extraction, and nobody creaming anything off the top. It's the exact experiment half my comments are asking for, and it's been running for 25 years. And the result? It was bought with £1.9bn of borrowed money, that debt has been rolling ever since, and researchers at Greenwich University found it still loses 41% of every bill to finance costs
 the same as Thames. Welsh customers took the largest single bill rise in the sector last year, and are projected to end up with the highest bills in England and Wales. If you take away the greed, but keep the same structure, you get the same outcome. The reality is... not one of the options proposed in the comments of the original post puts a single pipe in the ground, which is what all this is really about. The massive black hole is still there. It's £290bn deep. And as I'm sure you're beginning to realise, it can only be paid for by you
 through your bills, through increased taxation, or worse
 through the continued decay and disintegration of our water infrastructure.
I still don’t think many people have clocked how massive this story actually is. Thames Water is carrying £19.8bn of debt, up 2 billion in a single year, and it will run out of money before Christmas. £19.8bn against 16 million people who cannot switch, can’t leave, and cannot stop drinking water. That’s about £1,240 per head. 5 grand for a family of four. In some way, shape, or form, they’re on the hook for it. That’s the part that I don’t think has landed yet. This isn’t a story about sewage in the Thames, or bonuses, or another regulator caught asleep. The actual event is 30 years of a monopoly being used as collateral by people who knew its customers could never walk away. The bill has now come due
 and it’s a big one. The pipes and the infrastructure were never the real asset. The 16 million captive water drinkers were. What’s going to really sting is the fact there’s only two ways this gets settled. Your bill goes up, substantially, or your taxes do. Most likely both, and it’ll be on a schedule designed in a way so you don’t notice the hit, in an attempt to suppress the rage you should rightly be feeling. And before anyone tells me the creditors are taking a 30% haircut, look at what they’ve asked for in return. Fines waived until 2030. Pollution targets “significantly modified.” Bills raised above what the regulator allows. That isn’t exactly them eating the loss now, is it. That’s them buying a regulatory holiday, on debt most of them bought at distressed prices. Nobody voted for this, nobody borrowed it, and nobody saw the benefit of it. The debt was loaded onto a captive customer base over 30 years and paid out to shareholders who have long since gone. 16 million people are about to find out that they co-signed something they were completely unaware of. Now, this is what should worry us all. Thames isn’t a rogue outlier, it’s just the first one to fully hit the wall. English water carries north of £60bn of debt. Southern is already junk rated, needed a £1.2bn rescue from its shareholders, and its customers are looking at a 48% bill rise this decade before you count what the CMA added on top. Every one of these companies borrowed heavily when money was free and now have to refinance it all in a world where it isn’t, while being told to spend billions on infrastructure they left rotting for 30 years. Thames is just the first and most visible of what will likely end up being a line of dominoes. Maybe Burnham nationalises Thames Water. But ask yourself
 how many more will need to be nationalised? And who do you think pays for that?
25
52
9
180
36,368