ej88

surprised nobody responded with the most straightforward, occams razor explanation

they think what they're doing is actually good for society

not everyone is in the hackerspace libertarian / socialist sphere

i used to work for a place that used persona despite it adding extra friction to signups (literally resulting in less paying customers to the dismay of PMs) because it was worth it to combat fraud. theres a tradeoff in everything

Unfortunately agreed, i see swe (already) going the way of finance

most jobs are okay comp 9-6s with a small group of elite talent making extremely outsized comp

more emphasis on pedigree, exclusive hiring pipelines from top schools

Even the source article in the first link, https://www.nber.org/papers/w34836

the same firms "predict sizable impacts" over the next three years

late 2025 was an inflection point for a lot of companies

claude code 1B+ arr

ant 10xing ARR, oai

harvey legora sierra decagon 11labs glean(ish) base10(infra) modal(infra) gamma mercor(ish) parloa cognition

regulated industries giving these companies 7/8-fig contracts less than 2 years from incorporation

the top % of talent is still extremely hard to get, perhaps moreso

saw an article recently where every sector is seeing a reduction in IT/devs except for tech and ai companies

if your company is in a sector where eng is a cost-center and the product is not directly tied to your engineers / your company is pushing for efficiency it's an employer's market

We're already seeing eye-watering, blistering growth from the new hot applied AI startups and labs

Imo the wave of top down 'AI mandates' from incumbent companies is a direct result of the competitive pressure, although it probably wont work as well as the execs think it will

that being said even Dario claims a 5-20% speedup from coding agents, 10x productivity only exists in microcosm prototypes, or if someone was so unskilled oneshotting a localhost web app is a 10x for them

what doesn't make sense? "writing" the code is implementation

you still need good swes to distinguish if the generated code is good or bad and adjust the agent and plan the system

ime opus is smart enough to oneshot medium to small features by learning the existing codebase provided you give it the business context

I would posit another large factor is "owning" the software comes with the long tail of edge cases, bugs, support, on-call, regulations, etc... that an established SaaS has learned and iterated on for many years with many customers.

For the vast majority of companies they would (and should) rather let the SaaS figure that out and focus on their actual company

the perimeter of the tasks the LLMs can handle continuously expands at a pretty steady pace

a year ago they could easily one shot full stack features in my hobby next.js apps but imploded in my work codebase

as of opus 4.6 they can now one shot full features in a complex js/go data streaming & analysis tool but implode in low latency voice synthesis systems (...for now...)

just depends on how you're using it (skill issues are a thing) and what you're working on

"not enough people are emotionally prepared for if it’s not a bubble"

i am somewhat worried in the short term about ai job displacement for a subsection of the population

for me the 2 main factors are:

1. whether your company's priority is growing or saving

- growing companies especially in steep competition fight for talent and ai productivity results in more hiring to outcompete

- saving companies are happy to cut jobs to save on margin due to their monopoly or pressure from investors

2. how 'sequence of tasks-like' your job is

- SOTA models can easily automate long running sequences of tasks with minimal oversight

- the more your job resembles this the more in-danger you are (customer service diffusion is just starting, but i predict this will be one of the first to be heavily disrupted)

- i'm less worried about jobs where your job is a 'role' that comes with accountability and requires you to think big picture on what tasks to do in the first place

ai-first via mandate: throes of a company trying to appear cutting edge

ai-first via the actual technology being built: talent magnet despite the HN bubble appearing otherwise

the new meta is purchasing rl environments where models can be self-corrected (e.g. a compiler will error) after sft + rlhf ran into diminishing returns. although theres still lots of demand for "real world" data for actually economically valuable tasks

i mean llms

happy to be corrected but im not aware of any direct improvements llms bring to ultra low latency market making, time to first token is just too high (not including coding agents)

from talking to some friends in the space theres some meaningful improvements in tooling especially in discretionary trading that operate on longer time horizons where agents can actually help w research and sentiment analysis

oh, i meant copilot as in microsoft copilot in vscode. i havent used zed so can't speak to it but if it works for you it works!

copilot is a much worse harness, although recently improvements in base model intelligence have helped it a bit

if you can, use cc or codex through your ide instead, oai and anthropic train on their own harnesses, you get better performance

1. execs likely have spend commits and pressure from the board about their 'ai strategy', what better way to show we're making progress than stamping on some kpis like # of agents created?

2. most ai adoption is personal. people use whichever tools work for their role (cc / codex / cursor / copilot (jk, nobody should be using copilot)

3. there is some subset of ai detractors that refuse to use the tools for whatever reason

the metrics pushed by 1) rarely account for 2) and dont really serve 3)

i work at one of the 'hot' ai companies and there is no mandate to use ai... everyone is trusted to use whichever tools they pick responsibly which is how it should be imo

seems to be a pattern:

[Company that's getting disrupted by AI: Fiverr, Duolingo]: rush to adopt internal AI to cut costs before they get undercut by competition

[Company that's orthogonal: Box, Ramp, HFT]: build internal tools to boost productivity, maintain 'ai-first' image to keep talent

[Company whose business model is AI]: time to go all in

It's not like you need to take a course. The frontier models are the best, just using them and their harnesses and figuring out what works for your use case is the 'investing in learning'.

i would disagree with your dismissal of it as theatre. it's not forced, and it would probably be worse if I skipped those meetings :). meeting customers is not the same mental load as focused coding but it's still work that needs to be done

fwiw i think there's a balance here

at my current company i happen to work 70hrs/week but it doesn't feel like a ton of work, i'm having fun and let's be honest a chunk of the "work" is meetings & hanging out with my coworkers who are also my friends. the vast majority of people's productivity drops off after 4-6 hours of focused work. if i wanted to rest and vest there's plenty of companies to do that but your upside is capped hard

a company that 'requires' 996 doesn't understand why people work that hard in the first place.

I would add on that the most of the premium of a modern SWE has always been on understanding problems and systems thinking. LLMs raise the floor and the ceiling, to where the vast majority of it will now be on systems and relationships

the specific tasks (i.e. writing code) might disappear

but the actual work of constructing reliable systems from vague user requirements with an essentially unbounded resource (software) will exist

I'm quite convinced that software (and, more broadly, implementing the systems and abstractions) seems to have virtually unlimited demand. AI raises the ceiling and broadens software's reach even further as problems that previously required some level of ingenuity or intelligence can be automated now.

I would rather learn how to use these tools effectively now and ship more and of higher quality now.

If these tools improve to the point where anyone can pick it up - that's great! I enjoyed my head start while it lasted.

If these tools continue to require experience and a skillset to use, that's great too - I'll continue to learn and pull ahead.

Most of the gains come from post-training RL, not pre-training (OpenAI's GPT 5.2 is using the same base model as 4o).

Also the article seems to be somewhat outdated. 'Model collapse' is not a real issue faced by frontier labs.

I'm sorry your teammates have skill issues when it comes to using these tools.

I primarily find them useful in augmenting my thinking. Grokking new parts of a codebase, discussing tradeoffs back and forth, self-critiques, catching issues with my plan, etc.

We have a giant monorepo, hence the shallow clones. Each Claude works on its own feature / bug / ticket though, sometimes in the same part of the codebase but usually in different parts (my ralph loop has them resolve any merge conflicts automatically). I also have one Claude running just for spelunking through K8s, doing research, or asking questions about the codebase I'm unfamiliar with.

I implemented some of his setup and have been loving it so far.

My current workflow is typically 3-5 Claude Codes in parallel

- Shallow clone, plan mode back and forth until I get the spec down, hand off to subagent to write a plan.md

- Ralph Wiggum Claude using plan.md and skills until PR passes tests, CI/CD, auto-responds to greptile reviews, prepares the PR for me to review

- Back and forth with Claude for any incremental changes or fixes

- Playwright MCP for Claude to view the browser for frontend

I still always comb through the PRs and double check everything including local testing, which is definitely the bottleneck in my dev cycles, but I'll typically have 2-4 PRs lined up ready for me at any moment.