https://www.linkedin.com/pulse/leaked-tape-shattered-silicon-valleys-ai-illusion-jean-pierre-mugenga-fil9e?
9 replies (most recent on top)
@ea don’t forget the metaverse, Zuck spent billions on making people attend meetings in leg-less avatars 😭😭
@j4 You’ve never actually written code if you believe that.
What happens to all the frauds who have called themselves "AI experts" on DinkedIn though?
If Claude and Codex development stopped today and never progressed one step further, both would be the greatest coders in history, with no single human likely to even come within spitting distance.
Most devs are in deep denial. Business coding is not that hard. Most businesses haven't even really started using AI. By mid-2027 you'll see the rest of the SP500 companies finally wake up and start using this stuff.
AI might not be able to claim a Nobel Prize in physics for a few years, but it can most definitely write for loops faster than you, and most of what you do is write for loops.
US business coders are at a unique disadvantage. Pay for US business coders is historically higher than other regions. We are more vulnerable due to our higher costs.
I'm predicting a full on 50% reduction in global dev headcount by 2029. This will be one for the history books. That's assuming Claude and Codex and the free models only get slightly better than they are today.
The real value won’t be in models alone. Once this stuff moves into real enterprises, it becomes about data, security, observability, networking, silicon, and infra. That’s where Cisco has real leverage.
@ex I think you’re absolutely right. As it is today, there are some very good local models which are free to use. As the cost of using the paid models ratchets up, investors are going to be more vocal about their concerns about ROI. Once a major company makes a commitment to building their infrastructure to run local models on their own, I think the floor will start to fall out for the public companies. Hence, the rush on the part of OpenAI and Anthropic to cash out with their IPOs before the crash.
Anthropic and OpenAI are one good open source model away from being irrelevant. Then Infrastructure spend will collapse while personal compute will see a revival as highly distributed AI becomes the future.
It’s micro-computing vs mini-computing all over again.
This isn't indicative of a failure of AI as an industry, just an indictment of how broken Meta really is (despite the healthy earnings)
Everyone who gets deep in to that org says the same thing - Zuckerberg is surrounded by sycophants who validate him instead of challenging him. So bad decisions go unchecked.
Like throwing megabucks at Alexander Wang and expecting he could just conjure up an Anthropic inside Meta. It wouldn't surprise me to see Wang ousted by the end of the year.
No one doubts that with enough money, Meta can produce a decent model...but at this point they have to leap over OpenAI, Anthropic, Google and the free weight models by a substantial margin...no one sees that happening.
And now that Meta is thinking of turning their data centers into something like CoreWeave smells of desperation, but it also sounds like all Meta employees will just tell Zuck that its a "brilliant idea!!!!!"
@OP
Thanks for sharing. That Meta stuff is wild.