HearingTheQuantum @quantumpod

What he said

Director Michael Kratsios @mkratsios47

3 days ago

Read it at whitehouse.gov/science

27 53 398 85K 292

I recently scraped hundreds of arXiv physics papers for open questions, and fed them to frontier AI models. Here is what happened: m-malinowski.github.io/2026/06/05/ai-…

My new Substack, all about the key "aha moments" in the history of science & technology inventiontime.substack.com

So true

Adi @adi_baradwaj

11 months ago

every aspiring home cook has roughly the same journey 1) Basics with Babish 2) omg sous vide is amazing 3) why isn't home kitchen closed-loop industrial PID controlled cooking a thing?

1 0 12 1K 1

New Anthropic Research: Project Vend. We had Claude run a small shop in our office lunchroom. Here’s how it went.

@MichalOszmaniec @annawitten MS: "research published today in Nature, along with data shared at the Station Q meeting" = mamy artykul gdzie nie ma qubitow ktory mozesz przeczytac, i mielismy tez wewnetrzne spotkanie gdzie byly qubity ale tam ciebie nie bylo :)

Trying to forecast the future of quantum computing be like

I'm not yet convinced this exodus will last, but in case it does, you can find me at bsky.app/profile/msmali… But don't worry, I'll hang out here still!

We have a existential manufacturing problem in America. Why aren't there Amazon/Tesla-scale gigafactory warehouses filled with CNC machines? Will we ever be able to make our own products again? How do we accelerate the turnaround of American manufacturing? I asked this question a few months ago, and it's more relevant than ever. I firmly believe it's unacceptable to be a country that, ultimately, isn't able to make its own stuff. This isn’t just about economics—it’s about national resilience. It’s about the stability, defense, and sovereignty of the United States. A country that cannot produce its own goods is inherently fragile. Maybe this is irrational patriotism or stubbornness. Defense considerations aside, we must strive for price parity with our neighbors—subsidies can provide a short-term boost, but they are not a sustainable foundation for a competitive manufacturing economy. Future growth will require a return to cash-flowing, material-output businesses depending heavily on manufactured goods. Controlling our own destiny is a good thing. I’d like to kickstart a series of essays in coordination with @newindustrials, Blueprints for Reindustrialization, by contributing the first piece and setting the tone for what I hope will be a lively conversation from boots-on-the-ground operators. A few spicy takes up front. We’re not in first place. We killed the golden goose to increase quarterly profits. We lack both the capacity and the skills. We ought to create more high-paying manufacturing jobs, not low-paying manufacturing jobs. The most fundamental thing we need to do is increase efficiency. Manufacturing costs are inherently higher in the United States, period. After adjusting for the cost of living an average salary in Shenzhen is approximately $2,730, while in Houston, it's about $4,414. We’re 61% more expensive right off the bat. In addition to labor differences, other nations actively subsidize their raw material inputs. It’s tough to make the economics work when you’re so far behind. I’ve been on the front lines of making complicated stuff quickly for most of my life. The last four years of operating in defense-critical applications have been eye-opening. Whereas previously I could reach out to international partners for work, now I have to comply with U.S. rules for sensitive technologies. The contrast in capabilities and skill sets between the two locations is incredible. It feels like a barren wasteland of talent, with a huge vacuum left by the retiring skilled tradesmen. In addition to a population crisis, we are certainly facing a hardware-skilled labor crisis. Nearly every week, I’m desperately begging vendors to take our money. It's quite literally like the Futurama Fry meme—money in hand, begging someone to take it. More often than not, the answer is “we don’t have the capacity.” I’m a tiny customer, and we have no macro-scale event currently happening. This is not what national resilience looks like. Here’s an uncomfortable truth we need to say out loud: China has its shit together in a way difficult for us to replicate, in regard to manufacturing. They've made it a top-level priority to be the world's destination for manufactured goods. Moreover, did we think they would stop at dollar-store trinkets? Foolish. Every dollar of revenue in China is compounded into creating their huge critical mass of machines, infrastructure, and expertise. The plan has been running for decades now. They're climbing the ladder of technology as we paid them to do it. We could do the same. We could simply decide that it matters to make stuff here. It seems like we have a few knobs we can turn: • External taxes (import tariffs) • Internal tax breaks (low-cost/risk loans) • Technological breakthroughs (automation) • Operational excellence (scale) • Irrational actor patriotism Now, let's be realistic here. No amount of impassioned chanting of dynamism is going to shift the tides, though the X (formerly Twitter) grassroots discussion will kickstart the conversation. We're dealing with a scale truly hard to appreciate. We’re at least a trillion dollars behind. Policies to nudge this in the right direction are a bit over my head; I’m just an engineer with some ideas. I am, however, quite curious to sit down and understand what D.C. thinks about this. How do we foster growth and make domestic manufacturing powerhouses, akin to Foxconn and Pegatron? How do we pick winners and put the pedal to the floor? Before diving into specific scale considerations, let's address the macro issues. We need a set of policies that can accomplish several objectives simultaneously: • Capital equipment incentives: implement dollar-match assistance to incentivize companies to invest in capital equipment now, jump-starting the manufacturing sector this fiscal year • Long-term corporate tax breaks: provide tax incentives to corporations that successfully and measurably return their supply chain domestically • Future import tariffs: this is the “buy equipment now, China pays later” approach. Announce import tariffs 1-2 years in advance and use the proceeds specifically to defray the costs of upfront incentives. This gives domestic industries time to prepare and enhances competitiveness when the tariffs take effect. • Worker payroll credit: offer tax reductions to employees in specific manufacturing jobs (identified by NAICS codes), which appear directly on their paychecks. This boosts enthusiasm for the field and makes manufacturing careers more attractive. Capital Equipment Incentive. We should consider implementing a CHIPS Act equivalent for broader manufacturing—offering targeted, low-interest loans and dollar-matching on new capital expenditures. Imagine gov-backed loans at the Fed rate, paired with a dollar-for-dollar match on new machinery and automation purchases. This approach must be carefully executed to avoid turning into a grift-a-thon. The goal is to provide the activation energy for private industry to shift toward domestic manufacturing, not to heavily manipulate the economy or dictate specific actions. We should apply a light touch—enough incentives to motivate small businesses to expand in a healthy manner without distorting the market. If anyone on Capitol Hill is looking for bill names: • MARS (Manufacturing and Reshoring Strategy) • FORGE (Fostering Onshore Resilient Growth & Employment) • PRIME (Promoting Resilient Industrial Manufacturing Economy) Future Import Tariffs. The primary purpose of tariffs is to make domestic manufacturing more attractive by creating price parity between imported and locally-produced goods. If imports and domestic products cost the same, there's a compelling reason to choose domestic suppliers. Unfortunately, this approach often leads to higher prices overall. Announcing import tariffs two years in advance—a "buy equipment now, China pays later" strategy—can encourage immediate investment in domestic manufacturing. This gives domestic industries time to prepare, invest in capital equipment, and enhance competitiveness before the tariffs take effect. Importantly, the revenue generated from these tariffs could be used to fund the proposed capital expenditure (capex) incentive purchasing program. By directing tariff proceeds toward dollar-match assistance and tax incentives for domestic manufacturers, the tariffs not only level the playing field but also provide the financial means to strengthen the domestic supply chain. This ensures that the funding ends up in the right bucket, enhancing the effectiveness of the incentives and making domestic manufacturing a more viable alternative. It's crucial to remember that tariffs are only a tool—a means to an end, not a permanent solution. They should be used cautiously, much like steroids for domestic industries. In the long run, there are only two sustainable, tariff-free paths to achieving global competitiveness in American manufacturing: • Specialization: becoming the best in a specific sector through deliberate investment in capital equipment and the development of a highly-skilled technical workforce •Technology-driven productivity gains: achieving productivity improvements that offset our comparative labor and raw material cost disadvantages Achieving specialization requires conscious decisions and priorities. For instance, Japan's metrology industry exemplifies how a nation can become a leader in a specific field. Mitutoyo precision measurement tools are cherished by machinists worldwide, reflecting Japan's cultural emphasis on making tools as excellent as possible rather than merely adequate. While technology is often touted as the silver bullet, it's challenging to implement and defend over the long term at a national scale. Limitations like material science dictate the performance of tools—for example, constraints are more about the maximum surface feet per minute achievable by tungsten carbide than machine rapid speeds. Automation, although promising, has so far failed to deliver transformative gains. It often shifts costs rather than reduces them, moving expenses from production workers to capital expenditures and software development. Efficiency gains rarely outweigh these increased system costs. Numerous projects have attempted to replace a $15/hour worker with a robot arm, only to fail spectacularly. It’s a bit of a blanket statement, I admit, but is directionally correct based on my boots-on-the-ground experience and conversations. Moreover, technological advantages have a shelf life; they diffuse into the global market over time. A breakthrough might buy a decade or two of increased productivity, but resting too long on these gains risks falling behind. Unlike semiconductors, where Moore’s Law drives rapid advancement, manufacturing technology evolves more slowly, with hardware typically remaining viable for 10–20 years before becoming unprofitable. The reality is that deploying technology in this industry is difficult. While many small improvements are available, few shops have the resources to integrate these disparate pieces effectively. It's unrealistic to expect a typical machine shop to have dedicated software engineers to troubleshoot low-level FANUC communication issues with robot handlers or set up complex databases to coordinate RFID toolholder tags. One CNC shop owner shared a stark perspective: he believed there wasn't a single truly profitable shop in the U.S.—most were simply coasting on fully depreciated machines. While this view is extreme, it underscores the challenges and sentiments within the industry. Worker Payroll Credit. To attract talent and rejuvenate interest in manufacturing, I propose a worker payroll credit that offers immediate tax reductions to employees in specific manufacturing jobs identified by NAICS codes. This policy would directly increase take-home pay, making manufacturing careers more financially attractive. By targeting these critical sectors, the benefits reach those essential to rebuilding our industrial base. This approach rewards current workers and incentivizes new entrants, helping to restore the middle-class American dream associated with manufacturing jobs. Immediate benefits include: •Attracting new talent: higher net pay makes manufacturing competitive with other industries •Retaining skilled workers: financial incentives boost job satisfaction and reduce turnover •Stimulating economic growth: extra income leads to increased local spending and job creation •Enhancing respectability: recognizing and rewarding workers elevates the profession's status Administered through the existing payroll tax system, employers could adjust withholding based on the employee's NAICS code, increasing net pay without additional administrative burden. By providing immediate financial benefits, this policy affirms that manufacturing is essential to the nation's economic health and makes it a more attractive career choice. These policy measures lay the groundwork for revitalizing domestic manufacturing, but they are just one piece of the puzzle. To truly accelerate reindustrialization, the active participation of large corporations is essential. This brings me to the critical role of cost of capital and how leveraging the resources of cash-rich mega-corporations could be the game-changer we need. Cost of capital is everything. I still think one of the highest-leverage moves is encouraging the cash-flowing mega corporations in the U.S. to consider investing in domestic manufacturing capabilities. The ideas below are tailored for Amazon (@jeffbezos) but could be implemented by a handful of companies. We've also seen how one incredibly stubborn guy could stand up world-class manufacturing in the U.S. and turn a profit (@elonmusk). Like them or hate them, the billionaires serve an incredibly useful market function. You've got to realize that the activation energy to push this type of idea often doesn't come from a quarterly profit-driven market. Aggressive offshoring got us here in the first place. It takes a slightly irrational actor with a LOT of resources and a HUGE amount of conviction to make a move that pays off in ten years vs. the next quarter. I believe that measurable change is more likely to be affected by these entities than a group from YC. Sorry boys, but it’s a billions, not millions type of problem. Why am I specifically mentioning cash-flowing large corporations? Shouldn't I be advocating for the VC approach? No, in most cases. This is something I feel very strongly about. There are certain buckets of money for each type of problem. It’s important to know where it's appropriate to use high-risk capital (VC), profit-optimizing capital (PE), free cash flow, or government capital. I think that, net-net, VC money is not well-suited for the commodity manufacturing space. It comes at an extremely high cost and contorts business models into places where they won't operate sustainably. Risk capital belongs in areas going from 0-1 where there can be 100x–1000x improvements (someone should go fund a startup to develop next-generation tungsten carbide or single-crystal cutters). We’re talking macro level, and we need low cost of capital (Fed rate plus a percent) on the order of tens of billions. Okay, end of the Econ 101 section. I want to focus on actionable things that are more scale and operations focused. These can be realized with minimal (zero) breakthroughs. If it works, it would only be rendered more competitive and effective by other upstream policies. Efficiency is the silver bullet that makes everything downstream just work better. So let’s talk specifics. Let’s exhibit bias for action. What's actionable today? How can "unfair" advantages be leveraged today to save 3–5% in a bunch of small areas? Remember that I’m tailoring this to an Amazon-like entity; this isn’t for a 50-person shop. • Gigascale (>10,000 deployed machines) • Focus on operational excellence first, not technology • Don't invent anything new • Don't over-automate • Don't target high gross margin parts. Go med/low. • Utilization of equipment. Most shops are running 10–12 hours a day at best. Run 24 hours a day. • Own your freight network. The cost of moving material around is a significant percentage of your profit. Consider that most 3-axis CNC work comes in at 7x–10x the material cost, and materials average $4–$6/lb for nonferrous. • Access to cheap capital. Leverage your massive free cash flow to block-buy production from top machine manufacturers in exchange for large discounts. This is one of the biggest knobs, in my opinion. I know DMG will knock off 40% for a big enough order of machines. Many small shops are operating on machinery loans of prime plus 3–5% at MSRP. • Administrative support. Take advantage of your huge existing and battle-hardened accounting infrastructure. It's going to crush the one-person accounting department operating with QuickBooks. • Bootstrap on your own products. There are hundreds of millions of units of Amazon-brand consumer electronics already being sold. Every one represents a bill of material with at least a few parts well-suited for in-house manufacture: injection-molded cases for Kindles, microwave oven sheet metal frames, etc. • Massive deal flow. Offer deterministic pricing/lead times in exchange for being the default go-to vendor. You can operate break-even for the first few years to capture market share. At the end of the day, the engineers crave predictability and reliability. We're addicted to Prime 2-day in our personal lives—just imagine having 4-day turn times filled by your shop with 99% reliability. • Remote programming. Ironically, the most skilled portion of this job (the CAM tool paths) is the easiest thing to make physically distant. I'm typically a fan of being under the roof, but in this particular case, I recognize the difficulty of moving thousands of skilled workers to a few centralized hubs. With standardized cells/tools/workholding, it gets quite a bit easier for remote programmers to be effective. For quantity 1 parts, 80% is programming and maybe 20% is spindle time. Amazon already does Mechanical Turk. This is the skilled $100k+ version of that. I do this every day at my office; it's 100% possible. • Building space. Amazon's operational efficiency in deploying square footage is crazy. They've hyper-optimized the soup-to-nuts process of going from a plot of land to a finished and productive tilt-wall single-story warehouse. They already have a presence in every major city. •Software integration. Most shops lag far behind in software, relying on clunky, manual systems that can’t scale. Amazon’s expertise in inventory, tracking, and order management is a killer advantage. Streamline everything—raw material orders, production scheduling, and shipping. I was first applying these ideas to subtractive machining, but injection molding and sheet metal scale even better. They’re more deterministic, with simpler geometries and higher quantities. LEGO's Billund injection molding facility is a prime example of what’s achievable: over 1,000 machines running nearly lights-out. Apple is another excellent example, with the legendary football fields of Fanuc Robodrills and Brother Speedios knocking out watch frames by the millions. Yet, our current market setup is fragmented—96% of our manufacturing workforce is in small shops under 50 people. This decentralized model isn’t cutting it. When people say, "it’s too expensive to make parts here," what they really mean is that our output per employee is lagging. Let's review some estimates of the deployed machinery around the world to understand the league we're playing in. Even if we wanted to, we couldn't flip all of Apple's production to the USA—we simply lack the capacity. Subtractive (mills/lathes) •2,000,000 (Global) •300,000 (USA) Injection molding 1,600,000 (Global) ••145,000 (USA) Sheet metal •2,400,000 (Global) •365,000 (USA) The numbers are rough and pieced together, but directionally correct. The gap is staggering, and it's going to take years to catch up. As people have pointed out, there would be a huge problem of availability of the machine tools. Even Haas only produces ~2,000 machines per month. My friends in the industry are skeptical that such a thing could even be put together in a reasonable time frame. My experience has been 50-week lead times on transformers, 30 weeks to get high-pile racks permitted. While it seems physically impossible that @Tesla could build a gigafactory in two years, they accomplished exactly that. How do we remove some of the bureaucracy and roadblocks to infrastructure? Manufacturing is not glamorous, has never been known as such, and perhaps shouldn't be. But at least we could glorify it as a respectable and fulfilling profession. It’s dirty, hard work, and takes considerable amounts of effort to make significant improvement. For decades, software has been the title that works on complex and intellectual problems. Not only do we need to develop necessary skills and produce enough output for high-paying roles, it is essential that manufacturing regain its ability to provide a middle-class living. The gig economy is an embarrassment to our nation. I think we're in dire need of a skilled trade resurgence. The grand experiment of sending the entire population through college, only to exit with marginal real-world skills, has run its course. We need a pipeline of work that one can enter at an earlier age, gradually learn skills, and climb the ladder of the craft. It should take longer than a 4-week coding bootcamp to start a career. The post-war industrial revolution helped destroy the classic journeyman program in the United States. Other countries maintain working versions. Check out Germany's apprenticeship program (Ausbildung): • It typically lasts 2–3.5 years, depending on the profession • Apprentices split their time between vocational schools and practical training at companies • They can choose from 342 recognized trades • They receive a monthly salary and don't pay tuition fees • The curriculum is strictly regulated, ensuring consistent skill development across the country • Apprenticeships are governed by legal contracts between companies and apprentices. • After completing an apprenticeship, individuals can pursue further qualifications, such as becoming a master craftsman State of the market. Yes, I'm aware that Hadrian, Xometry, Protolabs, SendCutSend, and a dozen other smaller services exist. No need to blow up their Twitter handles; we all know each other. I have great respect for each of these companies. They each have their own particular niche they serve very well. • Hadrian has the biggest head start on the embodiment of a modern high-tech manufacturing process, attempting the monumental feat of end-to-end software controlled manufacturing. It’s the biggest swing by far, but will take an enormous amount of capital to scale to Foxconn levels. •Protolabs is the king of speed as an automated 72-hour turnaround time shop. • Xometry has instant pricing and a great marketplace for frictionless orders from real middle-America shops. •SendCutSend is a beloved institution for all things flat and bent, pulling off organic scaling with the lowest prices around. Each is a noble effort in trying to make us competitive, though I fear without government assistance in some form, these efforts are a drop in the bucket. The market has evolved into three or four distinct buckets of work. It all comes back to good, fast, and cheap. It's almost, by definition of the free market, impossible to do all three. If I were advising Jeff Bezos, I might suggest rolling up a few of the best players, but only the ones well-suited to be 50x in size. Honestly, the revenue ratios are just so low compared to any software M&A, everything in manufacturing looks like a bargain. Everyone is concerned with monopolies and antitrust in software; meanwhile, no one is looking at hardware. I don't think that a PE roll-up of existing small mom-and-pop shops would yield much, if any, improvement. Decentralized and diverse is the opposite of what I'm proposing. It's not a fully formed thought, but one I felt compelled to write down in between actually making parts. For the last two decades, I've been a consumer of manufacturing services, and an occasional manufacturer myself. I don’t have a book to shill, I just happen to have one foot in tech and one foot in blue collar. I started in a machine shop when I was 14, entering my mechanical engineering career with nearly a full apprenticeship under my belt. I started a hardware company in my 20s. Now I'm in my 30s, building hypersonic vehicles and working with America's greatest institutions. After thousands of parts both designed and made, I've had the chance to look around and consider the change I want to see in the world. • I want the U.S. to control its own destiny • I want skilled trades to flourish again • I want to see on-shoring/reindustrialization • I want McMaster/Amazon levels of speed and excellence applied to my custom manufactured goods • I want instant pricing and deterministic lead times • I want consistent quality. Doesn't have to be the best, just predictable • I want lifelong, well-paying careers for my younger brothers It seems like a fundamental shift occurred this month. How do we now translate that excitement into a plan? Rebuilding our industrial base is not just an economic necessity—it's a pivotal step toward securing our nation's future. The path forward doesn't require us to reinvent manufacturing, but rather rethink and optimize how we utilize our existing resources. By focusing on operational excellence, scale, and strategic investment, we can start down the long road of restoring America's position as a global manufacturing leader.

The sweet thing about a really snowy day is that people somehow start saying "hello" to each other when crossing on the street

@blip_tm Right! I'd basically want YC but staffed with a few chip veterans who can say: "this chip? no way", "that chip? only joe's foundry can make it, let me introduce you to joe".

Why can’t Britain build anything anymore? The news this week of the £100m ‘bat tunnel’ gave us some clues. Here’s the story of this tunnel, which has been 12 years in the making, and some thoughts on what it tells us

Interesting meta-analysis on Google's 2021 Nature paper claiming ML speeds up chip design: cacm.acm.org/research/reeva… TLDR: AI did not speed up the design, but it did make it slower 🐌 At least when it comes IC layout, humans still hold their ground 🦾! For how long though?

Does Zipf's law, or some equivalent, describe the distribution of the fundamental laws of nature? arxiv.org/abs/2408.11065 @Science_Cast @OxfordPhysics @NIST @APSphysics

I'll admit this would be fun. I used to get kicks from invoking the classic Ramanujan line. theorists: how did you get this pulse sequence?! and why does it work so well? me: It was revealed to me in a dream. [alas, they were not entertained.]

Agathatopian @MagisFuturum

2 years ago

@ma_sabba Or just add a note explaining the margins don't contain enough space to write it out.

0 0 2 1K 0

Quantum biology: where the very small meets life itself! 🔬🌱 Still amazed by Dr. @ClariceDAiello 's insights. Nature might be the ultimate quantum engineer after all! #MizterRadShow #FutureOfHumanity open.spotify.com/episode/2xRC8L…

It's statistical mechanics all the way down

Alan Baratz @Alan_Baratz

2 years ago

It’s incumbent upon everyone in the quantum community to avoid hyperbole and overstatements. We must stick with the facts and responsibly report on work being done without overstatements that could unnecessarily harm adoption and undermine the credibility of the industry.

4 0 19 8K 2

Some nice new opportunities available for quantum jobs in #quantum benchmarking and compilers from the Unitary Fund.Check out unitary.fund/careers/

Woah, this seems huge

Craig Gidney @CraigGidney

2 years ago

Been working on "Magic state cultivation: growing T states as cheap as CNOT gates" all year. It's finally out: arxiv.org/abs/2409.17595 The reign of the T gate is coming to an end. It's now nearly the cost of a lattice surgery CNOT gate, and I bet there's more improvements yet.

keyboard_arrow_left Previous keyboard_arrow_right Next

9 62 291 41K 86