AI Tech Layoffs Explained by ex-Meta Principal Engineer

It's greed period. META made 60 billion dollars last year. Appreciate the good thoughtful approach.

I was affected by recent layoff, I am now a full-stack Farmer

The astonishing thing in all of these cases is how little foresight corporate leaders have. Or perhaps how little care they have for upturning peoples lives just for another percentage point.

There are quite a lot of factors that you need to take into account which makes that diagram of 1 IC with 100s of agents not scalable: - Human intervention is still required. Can't tell you the amount of times I need to correct Opus 4.7 on a daily basis. Gets even worse if you downgrade the model because of the costs. - A single Eng. cannot switch context so easily to monitor just 5 agents implementing 5 different tickets. - Humans still generate and define the work to be done. That is a bottleneck too, you can't have 100 agents running 24/7 just because you can. - You still want engineers to keep institutional knowledge. It doesn't matter how well you document things, agents have limited context. This is the big question: what is the sweet point where I can still keep a lower number of engineers and assume the cost of using LLMs at the cost we have TODAY. Can you risk and fire half of your workforce hoping that the ROI will still be positive in 1 year, after LLM prices are still ramping up? Will open source model gets better fast enough? Not to mention the skill degradation narrative that is a fact when you just get lazy and rely everything to agents. Let me insist: frontier models are not perfect and hallucinate like hell. LLMs are a tool. Period. Yes, you can go faster, as we were faster when IDEs came out to improve coding productivity. This is just a next step in the journey.

Excellent analysis, thought provoking.

My theory is it was all part of the college industrial complex. When colleges became a business and was spinning up degree programs like underwater basket weaving so they can prop up enrollment and get students to sign up for exorbitant tuition with fed backing can’t bankrupt and cheap interest the only way students will continue to sign up is if there are promises of jobs. So companies did the hiring and students continue to sign up with the high tuition thus the ponzi. However you can’t continue to keep hiring and the ponzi scheme crashes thus the layoffs. The over hiring was bs, diversity manager, women in tech, program managers all fluff roles. Now with no fidelity in jobs for grads the colleges no longer seem like an investment and enrollment will go down. It was all part of the college industrial complex with collusion with corporations and the fed.

A reason for AI layoffs is that, during the intial implementation of LLM based AI it costs money so laying off relatively expensive employees that could be replaced by AI reduces labour costs freeing up operating cash to pay for AI tooling. This also motivates those remaining to adopt the tooling to managed their now increased work load. However, only in the past say 4 months or so have the LLM based models become good enough to substantially replace various white collar roles, though this has been happening for software developers for about a year.

Excellent analysis. This is exactly what we want. Thanks Kun Chen.

Excellent content man, wish you all the best in your journey

1:33-2:09 Esther Dyson: „They came, they surfed and then went back to the beach" I agree with you, and also work with ai.