🪴 15 startup/tech predictions for 2026:
Part 2 (nowhere near email length limit)
Part 2 of my predictions for startup/tech ecosystem going into 2026. Read part 1 here.
9) Accelerators finally stop funding AI-slop companies.
Please. Not another AI-slop IDE. Please.
One of the largest controversies YC had in 2025 was funding Chad AI (I refuse to link it, that's how trash it is).
Accelerators make bets. But when YC stamps their approval on AI slop, gambling apps, and other brainrot products, they're not just betting. They're legitimising garbage and telling founders this is the path.
YC didn’t become YC by funding everything profitable. They got there through conviction about what matters. Time to remember that.
10) Automating unverified tasks remains difficult but niche industries finally begin to solve this verification bottleneck.
AI crushes tasks where you can easily verify the output: math, coding, compliance. So what’s its bottleneck?
Everything else.
If you can’t verify a task, you can’t reliably automate it. That’s why AI capability feels uneven. Models look genius where success is measurable, and useless where “good” depends on human judgment.
All we need is to figure out how to “verify” a subjective task.
11) We're gonna experience more service outages, like we saw with Cloudflare. Twice.
The internet is held together by a few “pillar” companies.
AWS, Azure, Google, Cloudflare, Akamai.
These 5 companies hold over 70% of global internet traffic. When one goes down, half the web goes with it.
Cloudflare had two major outages in 2025. AWS went down for 15 hours, costing over $1 billion in direct losses. Vercel, same story.
More users. More AI traffic. More edge computing. Same fragile infrastructure.
We’re building faster than we’re building resilient. And every outage costs millions in revenue, trust, and productivity across thousands of companies who did nothing wrong except pick the same provider as everyone else.
The decentralisation crowd might actually have a point on this one.
12) Prediction markets like Polymarket and Kalshi double their market cap as degenerate betting goes mainstream.
The line between “forecasting” and “gambling” is getting real blurry, real fast.
Polymarket hit $3.7B in trading volume during the 2024 election.
Here’s my favourite picks for the bets that are live right now on Polymarket:
2026 is when your high school friend Carol starts saying things like “I’m long on a recession by Q3” while checking their Polymarket positions at brunch.
13) A new unicorn emerges by offering cryptographic proof to distinguish human-generated content from synthetic AI media.
We can’t tell what’s real anymore. Yes, that’s an actual problem.
Deepfakes are good enough to fool anyone. If you opened X over the course of the last month, most of the feed is “@Grok, put X in bikini”.
AI-generated photos, videos, voices, all indistinguishable from reality. Courts need proof. Platforms need verification. Brands need authenticity.
As of now, I don’t think there’s a cryptographic watermarking platform that actually works.
Big bucks if you manage to pull this one off.
14) SEO as a term finally dies, not that it doesn't work - no one even cares anymore.
If you still use Google to search for answers (I switched to perplexity, then to claude, then to something something and now am back to Google), here’s how it goes:
Search bar
AI summary (overview)
Show more of AI summary
One suggested article
“People also ask” dropdown
Sponsored websites
Actual SEO results
Mind this depends highly on the search, for some searches, e.g. product lookups the order differs.
So the question is whether it’s actually still effective to optimise for SEO to appear half the scroll-down button below?
What officially replaced it is GEO (Generative Engine Optimisation) and AEO (Answer Engine Optimisation).
But still the easiest way to have your product shown to users is to rule reddit and get shown in the AI summary as “one reddit user says".
There’s that one meme you probably know. Link is all you get, don’t wanna get flagged.
15) Defense technology companies replace FAANG as the most prestigious destination for the world's best graduates.
All I have to say here: have you ever seen anything sexier than Helsing CA-1 Europa?
How I can help
👨💻 Software
Need a top tech talent for your next product? Are you looking for your next gig? I’ve got you sorted.
📢 Advertising
Advertise in my newsletter to get in front of 11,095 founders.
🤿 Get more deep dives
Upgrade to a premium subscription to get an extra deep dive a week.
The verification bottleneck point is spot on. I ran into this when trying to automate content review at a previous gig, the challenge wasn't building the model but figuring out how to trust its judgement calls. Tasks with subjective quality metrics are where automation stalls hard. The shift from SEO to AEO is also underappreciated tbh. Most folks I talk to are still optimizing for search when the real traffic is coming from LLMs parsing their content and deciding whether to cite it.The reddit strategy you mention actually works tho, saw a B2B company double thier inbound by just being helpful in niche subreddits and letting AI summaries do the rest.