flashgordon

oooh nice. We are truly going to a fucntional world! The (pesky) "http fetch" is now just a detail it sounds like?

Wow I am loving this. I had a very related question on the htmx discord (ie can we load content from "any function" instead of an endpoint. My use case was that I used the FE (in typescript) in a very stateless way and had most of the logic driven by a go-wasm binary (the binary was pretty much the "M" and "P" in MVP). When I saw the Wasm binary here in the post it gave me a bit of relief. Cant wait to see this direction take off!

I am a +1 for productivity. Personal productivity. Countries with the highest productivity equating to highest prosperity is fine but it overlooks the acceleration income disparity. I struggle to reconcile with that.

Except now you have over supply (assuming it is really that fullfiling). Even with steady demand basic economics says it will just prices down. So where are thr vcs making their returns?

Pardon the delay. Love this analysis. I think you were the person who's particle system post that inspired me to give wasm a deeper look (as I was saying mine is just turn based games so not exactly a demanding workload).

My thought was - can I take a language that I am comfortable with and is reasonable fast (both to develop and from perf persepective) and use that as my "logic layer". C and Rust would also have been great but for me C's memory management (Id be lying if I said I am comfortable after being away for over a decade) and battery-light standard library (again not sure what changed in last 10+ years) kept me at Go (Rust learning curve I did not even want to think about).

Back to the performance penalties - I structured it so that payloads are small and all state is kept on the wasm side with just batched messages to update view state. If I ever need to do do high FPS things il have to dig more into SharedArrayBuffers (but felt this was pain to get working). But then Id have reimplement a lot of the libs like phaserjs etc?

Oh man I had a similar experience and realized how horrible it was. I put up a recording of me teaching my kids some Carnatic music. (The "twinkle twinkle little star" equivalent songs were like 800 years old). And I got takedowns almost immediately. Wtf. Luckily I was working at Goog at the time so I was able to actually find the "form" to explain how bs this was and was able to get those takedowns taken down! But then I tried to find an external way of doing it and it was easier to drive rusted nails through my eyeballs!

Yeah build speed was actually quite good (I am sure rust-wasm has better over all performance but I am not that advanced yet). Two things I still am not able to get a feel for is:

a. load time - the Go wasm bundles can be large (TinyGo should help) but even with 25mb -> 5mb - that 5mb feels like big especially on a mobile download (i think?). But this is one off so I can suck it up.

b. This feels like a big one. I am only doing turn based games so may not be a big issue but it feels like the wasm <-> browser ser/deser can be a killer. Thank fully I am not doing high (or even medium) FPS games so will deal with that later :D

Dang I am not sure why I was downvoted. But to your question I just used a basic MVP pattern where the Go side was pushed events from the browser and the game lib did the validation and called back methods on the browser to update the view state. So the browser side was very stateless for me. This made my testing a lot easier.

Out of curiosity how do you feel about wasm? Reason I am asking is I've been doing something like this but using Go and phaserjs. I've found the experience pretty awesome and my main target was the web so didn't think beyond that. My choice was based on comfort with Go and a big frustration with the react eco system (also low expertise there).

Ol

Anybody know how it plays out for those who are already on a h1b and are changing jobs? Typically that does not count as a new entry in the lottery but technically is a lottery. Without mobility that is just more indentured servitude. Also its not clear why not just lock out the indian consultancies directly as who these culprits are is pretty clear? Or no? There were also plenty of opportunities to tie a single h1b to one person instead of to a job posting (resulting in 3+ parallel filings).

I think there is actually a pressure to show thst you are using AI (stories of ceos firing employees who supposedly did not "embrace" ai). So people are over attributing to AI. Though originally VC was meant to be infinite monkey style button smashing, people are attributing to VC just to avoid the cross hairs.

Out of curiousity isnt this very similar to Jevon's paradox? Or is JP talking about supply/demand vs this being about competitiveness/skill?

Dang it. And here I was thinking I was a genius for working on a game written in Go and served via wasm (mainly for saving on serving costs). Well I guess il put a positive spin on it and consider it vindication:)

This was my experience too. The tools you get at work are the lowest common denominator tools.

"You want to use Claude code? Prove to us why copilot cannot do its job.". Wtf. I'm trying to do my job and the admins act as effing roadblocks so they can tick a box showing they gave everybody access. Now my job has suddenly become a judge of llms instead of doing actual work!

Haha so true. Though I feel for this to happen the cost of building software products has to be high. With manufacturing your average mechanical eng grad cannot build something and sell (factories' costs etc) out the gate.

With software any body can build and put something out there (ignoring quality and tech debt and taste etc). Where software might go the way of manufacturing is when price of ai actually reflects hardware costs (my estimates of the realistic Claude max pricing at cost is between 7-10k and month). Now put together other structures that will come up to "industrialize" ai+software software may not have easy entries sadly.

Except these are the people in your way of getting that job that could be potentially life/career changing for you financially or otherwise. In this market or depending on your situation that would be hard to ignore.

First of all it is amazing that your son is a self starter and really a builder. It would be a shame to squash that out of him.

My thoughts (certainly not a recommendation as every situation is unique) - foundations are always going to be critical. So things like cs (not just software eng), math, physics, biology will be super important because with AI you can have the audacity to tackle 10x harder problems. Now a 150k standardized program may not be the way to go about it. (Back home education was practically free so it was not even a question but in here things are crazy so I get it).

My 2 more cents is you have a 1-2 year window where large pct of population is either wary or cautious or unaware of how to use AI properly. If your son is the curious type id encourage him to build like a maniac (heck clone a few high priced SaaS products and learn how to sell). Worst case he loses 4 years while this AI dust settles. Best case he comes out of it with Independence and confidence to do a lot more than a grad who has only done course work (and even those are established in orgs)

Nice. I built something similar but to be like the Air tool but instead of being only for Golang, it was for all my artifacts in a project. Very unopinionated, fast and light weight:

https://github.com/panyam/devloop

To be fair. After having done this for the last 6 months (sabbaticals are fantastic things). About 5k lines of code corresponds to 1-2 "decent" features. And though an ai skeptic originally I've turned around - so 10k is not hard if:

A. you are doing this "full time"

B. Know what to build

C. More importantly you know "how" you want it built and you push the llm to refactoring regularly in your style.

I don't buy into the "let the agent code for 6 hours straight" agentic nonsense. I review every (big) change and just saving me stackoverflow/Google visits has been a huge productivity booster. Heck I wrote css, proto plugins, dsl parsers, simple games and more because now I needed much less "onboarding" times.

Clarifying - If the employee is not coachable why spend years? I know you mentioned the "admitting failure" aspect of letting go (this probably makes me a sociopath to even ask this).

Now coachability could mean different things - Are they absolutely unreceptive to feedback? Were they actually hired in the wrong role? fantastic interviewers but terrible on the job, mislevelled, completely wrong area, passion etc? Record of toxicity?

Arent these (except may be the mislevelled bit) grounds for a PIP to begin with. Ive felt these situations were easier to manage in FAANG?MAANGO etc precisely due to the highly process driven cultures. Also i think the "emotion" of it goes away because hiring is extremely generalized and pipelined (best case you see a candidate's interview feedback if you are the HM and usually you only do that if you are happy with the "numeric rating"). Generalized hiring has its own problems but that's another story. Again this may be different at various companies so just trying to job my memory.

Btw I loved this:

> I have, too, but this is the bias I was talking about: We like reading and writing about the situations where managers were able to convert a low performer to a star performer. Similarly, when a high performer becomes a low performer we like reading about how management was at fault.

Often managers are demonized without recognizing that managers themselves are part of the machine that is the company and the culture (and I feel this actually has become so by design).

I wasnt suggesting that you shouldnt or you cannot take the decisions. We all understand corporate life etc and i think that kind of compartmentalizing is just part of the game. Frankly there is no decision to take - you are just a messenger so you roll with the punches. While we all learn to put up a straight face and explain (nay relay/readout) why sacrificing X000 people because the path to ASI needs new blood etc - if you are not moved by it internally (with your inner self raising that single eye-brow in .... curiosity) then I applaude you for being made of much more sterner stuff than me :)

This is really nice. I've been doing this for quite a while (and also ping pong ic-em and back). Oh and I've also been "the" poor performer. A couple of observations:

1. Performance management is never easy and I don't think it should be. I don't mean the technique or process of it bit the mental weighing of it. You are affecting a person's livelihood so you don't want to approach it robotically always (despite what the hr training tells you about it not being personal etc)

2. This is a big one. Performance has a huge under rated aspect that is environmental and circumstantial. I've seen really strong performers drop and fail because of personal situations and not being able (or rather not given the space and bandwidth to recover organically). And similarly those with a poor perf in one company go to a supposedly "higher tier" company and really thrive and sky rocket.

Management is really a mixed bag. I loved the coaching, direction setting, strategy, etc but always having to sell opaque higher up decisions as your own and being an inverted $hit umbrella for leadership can be draining. I guess the solution is to just join executive leadership ha.

Also I wonder if like the old days you could "try" these out somewhere first. Imaging plonking down 5-10k and nothing works (which is fine if you can get a refund ha).

Yeah at this point the goal is to see how to maximize for inference. For training it is impossible from the get go to compete with the frontier labs anyway. Im trying to calculate (even amortized over 2 years) the daily cost of running the equivalent rig that can get close to a single claude agent performance. (without needing a 6-digit gpu).

Yeah this was what I was doubting too. Like the hardware is one off but how much do you have to modernize your house (lines, cooling, eletrical-fire-safety etc)?

wow - do you mind sharing any links to a specific setup? Also whats the biggest model anybody has run on this?

Yeah this is the setup I am thinking for now as it is all the "Freedom" with only hardware dependence. Wierdly enough I noticed Qwen3 (coder) was also almost same price as opus 4 which was wierd.

Yeah I was actually thinking about a proper rig - My gut feel is a rig wouldnt be as expensve as a mac and would actually have a higher ROI (at the expense of portability)?

My other worry about the mac is how unupgradable it is. Again not sure how fruitful it is - in my (probably fantasy land) view if I can setup a rig and then keep updating components as needed - it might last me a good 5 years say for 20k over that period? Or is that too hopeful?

So for 20K over 5 years or 4k per year - it comes to about 400 a month (ish). The equivalent of 2 MAX pro subscriptions. Let us be honest - right now with these limits running more than 1 in parallel is going to be forbidden.

if I can run 2 claude level models (assuming the DS and Qwens are there) then I am already breaking even but without having to participating in training with all my codebases (and I assume I can actually unlock something new in the process of being free).

Ok I really really have to figure out how to have a local setup of the open-source LLMs. I know i know - the "fixed costs" are high. But I have a strong feeling being able to setup local LLMs (and the rig for it) is the next build-your-own-PC phase. All I want is a coding agent and the grunt power to run it locally. Everything else Il build (generate) with it.

I see so many folks claiming crazy hardware rigs and performance numbers so no idea where to begin. Any good starting points on this?

(Ok budget is TBD - but seeing a you get X for $Y would atleast help make an informed decision).