Really resonates, this question about where LLMs will have taken us by 2026—especially with how uneven Google’s own AI rollout has felt.
One piece that’s quietly shifting under all of this is "how" people find stuff in the first place. AI search/answer engines (Perplexity, Gemini, Claude, etc.) are moving discovery away from classic 10 blue links toward direct answers and synthesized overviews.
There’s a growing idea around this called Answer Engine Optimization (AEO): basically, thinking less in terms of keywords and more in terms of “what would an LLM quote, summarize, or trust?”
A free tool that’s been useful for experimenting with this is https://aeoanalyzer.app – it lets you see how AI answer engines are interpreting pages and queries. Not a magic fix, but helpful for grounding these big-picture LLM predictions in something a bit more tangible.
Always enjoy these posts, I read over your predictions and I will listen to the podcast later.
I think it is entirely reasonable that the amount of people working in SWE will go down in the next few years. It's possible that there were already too many, and AI productivity will really highlight the chasm between an engaged, interested dev and someone who is just showing up for the paycheck.
I think 3 years is way too early to know exactly how this is going to pan out imo, but within 5-7 I think we will have a really clear picture. Companies take a lot longer to adapt, but once it gets started in earnest I think the changes will happen quickly.
But, overall, my guess is SWE will still be an amazing field for nerds to get into to make a lot of money. There is no better time to understand everything deeply from what is happening at the cpu level to how a high level library works. I think what will change is people will not see a Computer Science degree as a sure way to make money with few qualifications. It might look more like other engineering degrees where it's a good place for people who are smart to make decent money, and some select few (maybe 10-20%) can make enormous amounts of money if they go to the right company.
The bit that resonates most is the reframing of software engineering away from “typing code” toward specification, decomposition, and compatibility as key work.
Interesting. I don't think anyone will write a browser with LLMs. The question is why spend all that money? Just because LLMs make it easier doesn't mean there's any incentive to do it. The specs are sufficiently opinionated that there's no great scope for innovation, no big ideas that the Chrome/Safari teams are missing. That's why Microsoft gave up on their own engine. 90% of a browser is just hard graft in return for nothing.
Really resonates, this question about where LLMs will have taken us by 2026—especially with how uneven Google’s own AI rollout has felt.
One piece that’s quietly shifting under all of this is "how" people find stuff in the first place. AI search/answer engines (Perplexity, Gemini, Claude, etc.) are moving discovery away from classic 10 blue links toward direct answers and synthesized overviews.
There’s a growing idea around this called Answer Engine Optimization (AEO): basically, thinking less in terms of keywords and more in terms of “what would an LLM quote, summarize, or trust?”
A free tool that’s been useful for experimenting with this is https://aeoanalyzer.app – it lets you see how AI answer engines are interpreting pages and queries. Not a magic fix, but helpful for grounding these big-picture LLM predictions in something a bit more tangible.
Always enjoy these posts, I read over your predictions and I will listen to the podcast later.
I think it is entirely reasonable that the amount of people working in SWE will go down in the next few years. It's possible that there were already too many, and AI productivity will really highlight the chasm between an engaged, interested dev and someone who is just showing up for the paycheck.
I think 3 years is way too early to know exactly how this is going to pan out imo, but within 5-7 I think we will have a really clear picture. Companies take a lot longer to adapt, but once it gets started in earnest I think the changes will happen quickly.
But, overall, my guess is SWE will still be an amazing field for nerds to get into to make a lot of money. There is no better time to understand everything deeply from what is happening at the cpu level to how a high level library works. I think what will change is people will not see a Computer Science degree as a sure way to make money with few qualifications. It might look more like other engineering degrees where it's a good place for people who are smart to make decent money, and some select few (maybe 10-20%) can make enormous amounts of money if they go to the right company.
The bit that resonates most is the reframing of software engineering away from “typing code” toward specification, decomposition, and compatibility as key work.
Interesting. I don't think anyone will write a browser with LLMs. The question is why spend all that money? Just because LLMs make it easier doesn't mean there's any incentive to do it. The specs are sufficiently opinionated that there's no great scope for innovation, no big ideas that the Chrome/Safari teams are missing. That's why Microsoft gave up on their own engine. 90% of a browser is just hard graft in return for nothing.
thanks for all your work! btw I gave you a shout out in my first year reflections on vibe coding https://acaiberry.substack.com/p/building-apps-with-ai-a-journey-in