If you have a kid who is a recent software engineering graduate and you are as excited as I am about the H1B scam finally ending, here's how it started... IBM is almost singlehandedly responsible for the outsourcing of American engineering jobs. It was the early 90s and IBM was getting nuked by Microsoft, Oracle, HP, Sun, Dell, and Intel. Massive losses and the existential dread of a dinosaur monopoly being devoured by younger, hungrier, more nimble beasts. Using it's massive political clout and governmental relationships starting from all the way back in WWII, it smoothed the path for the 1992 partnership with TATA, the first and greatest tech outsourcer. Indian based employee numbers surged from 9,000 in 2003 to 74,000 by 2007, surpassing U.S. headcount. By 2006, IBM became the multinational with the most employees in India and by 2017 it overtook the U.S. as IBM's largest workforce location, with ~130,000 employees (one-third of global total) and today, IBM India employs ~150,000–200,000 people across 20+ cities. Once IBM opened the floodgates and smoothed the political/legal landscape everybody in the Valley copied them, with Microsoft being the most aggressive. Think about it: billionaires running our most profitable companies have been gradually destroying our scientific and engineering edge by hiring indentured servants instead of American graduates so that they can build a bigger yacht. Our elites literally sold out our children's future to make more money. But wait! It gets worse! IBM also spearheaded the massive expansion of H1Bs so that, as essentially an Indian company at this point, it could import a captive workforce where on-prem was required. The follow-on effect was the creation of a local infrastructure primed for colonization. Check out the leaders of the massive number of Indian owned IT firms around DC, and they're almost always IBM or IBM adjacent. This is especially true of the older ones, while the newer ones are often nepobabies of former founders. It's a tikka masala mafia. A personal aside: I have worked at Indian dev shops and with Indian partners and colleagues for decades. The caste system and the entrenched nepotism are 100% real things; it's literally how their culture functions, and it's why India itself is a non-functioning country. Also, I would not describe them as a particularly honest people. I can't count how many times I've been lied to by somebody with that weird bobbing head thing while they cheerfully smile. And the weirdest part is that they know that I know that they are lying, but they're lying and smiling anyway because they're not ashamed of doing it. There's a sense that if you are dumb enough to get cheated then that means that you deserved it, and it's an incredibly toxic and anti-western way to do business. Then McKinsey (those bstrds) took the model to Walmart, GM, Verizon, Fannie-Mae...everywhere. Very often in my career I've been like: "Wait: Am I even in America right now?". Once, I even got hired because they explicitly needed a white engineer to attend client meetings, while "my" entire team was Indian and half H1B. Nerd Aside: Remember IBM's UML and Rationale Rose, etc.? Those were created to communicate bugs and defects from US based analysts to Indian bases devs. IBM literally created the infrastructure of tech outsourcing, destroying (or never starting) the careers of American engineers. All so that they could make money selling a mediocre product using 3rd world labor. It's FN treasonous. In the parlance of Creighton's excellent Airframe, IBM "gave away the wing"

But honestly? The hardware was seriously compelling. I’m not one to swoon over Microsoft products, but I’ve also mentioned it a dozen times to folks as the shape of the future. So… watch this space?

Given that agents still need high level supervision. And given that the human review of code is a bottleneck. And given that the models have gotten good enough that strict deterministic harnesses are suboptimal. Then how should we supervise their output. My current solution is to use an interrogation tool coupled with an agent. The tool displays the structure of the system in a UML-like diagram, color coded with crap and mutation scores. The diagram is dynamic, allowing the user to click on element and drill down through the architecture -- all the way to code. The user can detect structures that need improvement and tell the associated agent to make the appropriate changes. These changes can range from high level architectural changes, to the lowest level code refactorings. The associated agent will make those changes and cause them to be redisplayed. The tool I'm using is custom made for me, but you can point an agent at it and use it as a template to build your own custom tool. It's a work in progress, so don't be surprised if it changes a lot over the next few weeks. But then you probably won't care because by then you'll be deep in customizing it for yourself. And anyway, you can always point an agent back at mine, or someone else's, and integrate improvements into your own. github.com/unclebob/uml-view….

"No matter how strong open-weight models get, most inference will always happen in AI datacenters." seangoedecke.com/local-model… < great hot take on why "local models won't win." It's not a zero sum thing. Open/local will have a great place in the spectrum of model options.

We're thankful to have you as our customer, and appreciate the feedback you've shared on our main cabin seating configuration. One of the great perks of being a status member of your level is the option to select Main Cabin Extra seating at no added cost.

Amen. I remember how sad I was to move away from the Bay Area. Thought life had ended bc I was no longer in what I'd believed was the center of the universe. Once outside it was like my brain rewired and I could finally see that most people live happy lives that aren't fixated on tech. It seems like such a silly thing to say but in the Valley's reality distortion field you come to think that there really is no other life beyond RSUs and outsized payouts.