leoqa

The cynical take is that with US companies expecting productivity increases via AI, they need to protect the US workers from competition via foreign labor. The current administration was voted in with an anti-immigration mandate so this is consistent. The practical reality is that you are not safe on any visa, it can be terminated arbitrarily by the state department and your recourse is likely expensive and timely.

Unfortunately the benefit of TLA+ is the act of modeling your system painstakingly. The actual checker helps confirm your hypothesis, etc. But skipping the modeling and outsourcing it is not ideal. I’ve always struggled reasoning about models my team mates wrote, and will often have to mentally go through the process of arriving at the same abstractions/invariants etc before I can understand it.

The higher velocity ends up bottlenecking on actual product decisions, deployments, testing etc. Before AI was generally blocked on the design approval, PR cycles, flaking tests etc. AI just helps me endure the pain of legacy code easier.

In my own personal projects I’m flying with AI. I know what I’m doing, I know how I would implement the code. Now I can just save the labor of typing the boilerplate.

People are arguing about the role of these HFTs being a net good etc. They’re missing the point, these bright kids are trading something more profound- a sense of purpose, a higher calling or passion- to simply run adversarial arbitrage and pump their egos up with puzzles.

I’m in big tech and use AI extensively, namely to do the same amount of output but in 1-2 hours a day. Been spending a ton of time on my side projects though.

Seems like a lot of energy dissecting C-suite news clippings. The reality is that no one cares about alignment and it’s only controversial for the naive or the dramatic among us.

Claude is useful for software engineers. It’ll be useful until something is better-enough and then we’ll all move on to that.

Most folks are using both Claude/Codex together anyways, undermining the idea that Anthropics corporate strategy mattered in the market.

He’s wrapped his identity in being a founder so can’t see the sacrifice/meaningless of it all.

Is the stakes and the ownership. 5 people in a coffee shop working on a 0-1 problem is a lot more stimulating than 90 people on a team shipping incremental updates to a legacy system.

This was always my least favorite part about being a software engineer (and the downfall of many): being a software engineer has become an identity crutch for many. I’ve see so many kids whose whole identity is being good at computers; they go through middle/high school, college getting affirmations about their value and intelligence. Then they get out of Stanford, Berkeley and show up to the feature treadmill that must keep moving but is weighed down by the 10000 short cuts made by the people who came before.

They burn out, or worse become toxic, because their shallow identity led them down the path to being a “Real” engineer and at the end of the day we’re not actually participating in any sort of real value creation beyond attention monetization.

The mystique wears off quickly and they don’t have real hobbies or interests, they basically talk about RSU packages at lunch and the latest tweets etc. I used to joke privately because almost every time we had lunch they spent most of the time discussing the optimal path to walk.

It’s unique in some way- you can’t be a good doctor or lawyer in middle school and the value system is geared towards maximizing paychecks and working in big tech. Once the reality sets in that you’re going to be doing sprint planning + standups for the next 20-30 years it can be a weird shock.

My first job was at a FANG and I lasted about 2 years- I remember riding the escalator in and seeing how miserable everyone looked on my first day. As an eager junior I reached out to the principal engineer in my org for mentoring, asking him what I could do to be better, faster. He told me: “go find a wife and don’t worry about work- you’ve got a long time left”.

At one point I looked at the senior guy running sprint planning and realized I didn’t want to be him. I bought a 1 way ticket and put in my 2 weeks. Went on to backpack around for a year then ended up at a startup where I made a bunch of friends working on real problems.

My hot take: reviewing code is boring, harder than writing code, and less fun (no dopamine loop). People don’t want to do it, they want to build whatever they’re tasked with. Making reviewing code easier (human in the loop etc) is probably a big rock for the new developer paradigm.

HSI was primarily the main investigative body responsible for human traffic and crimes against children prior to this administration. The second largest federal investigative agency behind the FBI (6k agents). Now doing immigration enforcement.

This is clearly another LLM response bud. Stop using it to communicate it’s too obvious.

Durable workflows are just distributed state machines. The complexity is there because guaranteeing a machine will always be available is impossible.

What’s up with all these bots posting 3-4 sentence summaries in the comment section?

Yeah I get the sense that terraform change application is solved by just serializing all changes? The concurrent applies isn’t that big of a deal?

I’m also fine with git, and have used mercurial and p4 before. I think simplicity is better in this case. I do think with more and more generated code inflating the codebase with high velocity, we need to find a better way to merge conflicts.

I was entertaining an offer from Discord and also stumbled upon the founder’s former company debacle. The platform vision pitched to me in the interview seemed similar and seeing as how he started to implement spyware I decided to bail.

You’re probably correct (I’m 30 and don’t have a policy) but I can’t think of any financial instrument that would guarantee skilled nursing ($4-8k/month in MCoL) for 10-20 years. No one in the bottom 90% of American earners can afford that drawdown even if they saved considerably.

I’m not particularly sympathetic to retirement home narratives that paint each person as a lonely, incompetent senior wasting away. I witnessed first hand how my mother spent her adulthood partying instead of providing and now it’s become a critical issue for our family to provide for her. Ultimately I don’t expect my children (or staff-journalist) to project pity on me for my circumstances.

I run a weekly route with MealsOnWheels and deliver food/perform wellness checks to many people whom are homebound. There are much worse fates for seniors than a community home with social programs and meal service.

I think it is everyone’s duty to buy long-term care insurance (perhaps the government should provide this, as we will all need it). I also believe you must provide for the retirement you expect.

Wait until you hear about Gummy Bears.

I quite frankly will just read the code. Go generally discourages abstractions so any code you jump into is fairly straightforward (compared to a hierarchy of abstract classes, dependency injected implementations, nested pattern matching with destructuring etc etc).

Regarding your IDE issues- I’ve found the new wave of copilot/cursor behavior to be the culprit. Sometimes I just disable it and use the agent if I want it to do something. But it’ll completely fail to suggest an auto complete for a method that absolutely exists.

I really disagree. I think hentai is a flimsy wrapper around child fetishization and needs to be heavily regulated. I think having rape or torture simulators are extremely harmful in multiple ways.

Really would prefer the government outlaw these things but I don’t mind companies protecting themselves from liability.

Usually this is the “leads the function but we’re not going let them into the c-suite”. The head of engineering gets usurped by the outside cto hire and put out to pasture.

The use was audited and is now being investigated. The claims were for various local and federal investigations. ICE also contains HSI, the second largest federal law enforcement agency, which prior to their recent mandate has been tasked to solve sex trafficking, import fraud etc. SF has multiple large inter-agency task forces that run multi-year long investigations into all types of crimes. HSI is part of those investigations. Querying flock to establish a suspect’s presence during the commission of a crime seems like it’s within the bounds of reasonable use.

Yeah people fail to provide examples but continue to be doomers about how easy it is.

Absolutely agree. I think the tactics and strategy are to blame here but ultimately I don’t think the administration is doing anything illegal. All cases I’ve seen where there was an alleged violation of due process were simply accelerated asylum denials with immediate deportation orders. I’m sure there have been some though, but I know local police, HSI and ERO folks that really don’t want to arrest non-criminal laborers yet are just following orders from the admin. I would be highly suspicious of any conspiracy that federal law enforcement agents are committing illegal arrests.

If it were up to me, I’d be pursuing some type of visa reform for permanent laborers that grants amnesty for those here with American citizen family members (i.e. birthright kids) as of a certain date. I’d make the asylum process occur outside of the country, signing agreements with the originating countries like Guatemala, Honduras to provide housing and food for asylum seekers while they pursue a claim.

I always like to frame it this way: ask someone what a reasonable response would be if they flew to Paris and then decided they didn’t want to leave. What is the French government allowed to do in their moral framework to enforce their immigration laws.

People don’t have a great answer. The asylum process actually works- it just turns out that many, many cases aren’t valid and it was abused to gain entry once we allowed asylum seekers to remain in country.

Why is this better than the other one?

Aside: I learned the Sep Axis Theorem in school and often use it for interviews when asked about interesting algorithms. It's simple enough that you can explain it to non-technical folks. "If I have a flashlight and two objects, I can tell you if they're intersected by shining the light on it". Then you can explain the dot product of the faces, early-exit behavior and MTV.

I actually modified SpiceDB to inject a groupcache and Redis cache implementation. My PoC was trying to build a leopard index that could materialize tuples into Redis and then serve them via the dispatch API. I found it easier to just use the aforementioned cache interface and have it delegate to Redis.