chris_money202

It depends on what you're doing. None of their models are Opus level, but not everything needs Opus. Their models are targeting cheap and useful for some common things not expensive and useful for anything

Nice didn't know that. I was thinking lower APIs similar to directX for gaming

Microsoft is doing things differently with AI. It feels to me they are moving into local inference heavily and see a future where Windows has native AI APIs that run locally or optionally in the cloud/edge.

The laws are outdated, tariff laws were written when trade was physical goods. The US heavily trades in services present day. But I don't need grounds because the conversation has been about establishing said grounds.

That is the planet Claude is from

I vaguely recall there being a semi universal remote shortcut (pressing stop, stop, play maybe?) that would skip right to the start of the movie. Was a game changer when that was discovered

Hey man I'm just a software engineer now, I can't speak on public policy.

My best effort, the contract is a red herring, we only care about the company that employs the abroad remote employees which is the contractual company in this case. If that contractual company is also abroad, we tax (tariff?) the contract.

That's true. Problem is a lot of the time Middle America is comfortable and doesn't want to move out to the valley or Seattle.

Remote work makes that easier now, but with silicon / EE its not always the most optimal way to work.

Middle America EE is also not building data center class chips or infrastructure most of the time, would it be easier to get one of them up to speed, yes. But it's the equivalent of throwing a front-end engineer into C or embedded programming or vice versa.

They may. I think the government should probably tax the diff in salary in those cases. The problem is, that is probably incredibly hard to define and pass as legislation or whatever has to be done.

H1B and offshore jobs can be beneficial to America though, and that was why the program started. Sufficient guardrails just never went up and now there is opportunity for some abuse. But that abuse isn't across the board, there is still beneficial H1B and offshoring going on mixed in with abuse.

Microsoft employs more than Software engineers.

Its Azure Hardware division that manages its first party silicon and its deployment has a huge hardware and electrical engineering presence. These fields don't have as many people in America in them, because Software has paid so much more for the last two decades, why would anyone go EE when they could go CS and make 2-3x more. In other countries, these degrees were paid similar so there is a larger pool to pull from then in America.

You can't train a software engineer overnight to be a silicon engineer

I fall into the camp where I agree with these antipatterns, so I'm probably not the target audience for the question. But a scenario could be if I was writing to appeal to PhDs or researchers, maybe I should include the links or citations on top of providing a brief summary of what a firewall is. Firewall might be a simpler example, but let's say a semi-novel algorithm for research sake.

My thinking there is PhDs/researchers a lot of the time like to dig deeper and are also pretty skeptical of new information and want to see it backed up by someone else unless the claims are completely novel, then they want the raw evidence typically in line or near what was stated as novel.

I think the biggest anti-pattern is not writing for who you are targeting, your intended audience.

I think your anti-patterns apply very well to the stereotypical SWE or HN reader, and if that is who you are targeting, you should absolutely follow these. But the biggest anti-pattern of all is following these anti-patterns while hoping to appeal to the untypical SWE or HN reader.

It comes down to your target audience. Everyone has preferences on what they read, someone who prefers formality might not enjoy casual blogs written how people talk. And vice versa. Software engineers in general probably prefer casual blogs, but there are software engineers that hold PhDs and prefer some rigor in the writing too.

I mean 1) hell yeah that's sick and 2) the front-end software isn't the moat typically, it's the cloud service with stuff like collaboration and storage, maybe even render?

Why is there any point in trying with faster memory if you are only at 80% perf on DDR3? You'll just go to 40% perf on the next iteration?

You should only upgrade when the bottleneck becomes memory, right now the bottleneck is still on compute until you are at 100% max perf

Tens of millions of dollars per AI agent isn't a blocker? In many orgs its hard for humans to get their hands on the proto equipment for long enough to run their tests.

Its a SYSTEM on Chip, evaluating 1 function on performance is superficial. This could have the best performance in the world, and it doesn't matter if the bottleneck is upstream

This is the smallest unit of a typical AI ASIC, for example Google's TPU would have several dozen more compute units inside of it per chip.

In essence this is the simplest unit of an entire AI chip. The more complicated units of AI ASICS are actually the periphery, especially around PCIe and Ethernet and the sub-systems that link many AI ASICs together to move huge amounts of data around ultimately to each TPU.

So its missing ALOT

There doesn't exist a single FPGA that can fit an entire AI ASIC. You would need dozens stitched together, then comes the issue of clock speeds, FPGAs typically run far below reference. There also memory issues with FPGAs.

Companies typically combined multiple platforms together such as HAPs, Zebu, Palladium, fleets of FPGAs, and Virtual Platforms in order to design and verify ASICS. So, AI would need access to tens of millions of dollars of HW and Software in order to build and verify a chip design.

That is probably true, but the ones that make it through to become experts might get super-charged with AI as well

"Solo Gambling" vs "Social Gambling"

Some people are comparing banning gambling to banning alcohol or tobacco.

Many of the ideas that later became the United States were discussed in a bar while drinking, the same cannot be said anywhere for solo gambling. Whether you like it or not, alcohol and tobacco are woven into the social order.

Well its not just any old human doing these things in a general sense. Its typically academics or highly paid researchers who love doing work like this. So, I don't think it will just one day stop

Another comment said treatment needs to begin 24 hours after infection or its almost always fatal. And they also said you don't have many symptoms within the 24 hour window.

I could see that still killing a lot of people even though a treatment would guarantee survival

How are they going to compete with Asus, Apple, Lenovo, Dell, HP, any other OEM?

Making a computer is about supply chain as much as its about what's inside and the software. Those supply chains take a decade to establish

It’s called design verification, formal proofs happen mostly at the EDA tool level and largely already automated. Design verification focus on functional correctness of the chip for its intended use case

I've heard a lot of stuff like people don't WANT to drink anymore but I wonder if it's not so much people's perception of drinking changed as opposed to people are in social settings where drinking occurs less. Alcohol is an addictive substance and best way to get someone hooked is small, frequent events where drinks are consumed. Over time they consume more and more and seek out more places to drink (and socialize) as the hook sets in. Peer pressure also plays a big role here too.

Wondering with remote work, remote school, etc post covid if we are just seeing less hooks set and less peer pressure.

And maybe it's just the harness and the backing model is mostly the same. But I would expect Codex to give better responses to code questions than chatGpt using the "same" model.

I went back through Azure foundry, and it looks like they have gotten rid of reasoning variants, so might just be one model now.

I'm pretty sure there is a completely different model for Chat vs Coding.

Like if you deploy 5.6 on azure foundry there is a chat model and a coding/reasoning model

Feel like this has always been a problem, especially with large legacy codebases. Very few people ever knew everything. We just reach the problem faster now; it used to take years of churn and turnover.