~/ reece-martin

2026 Predictions

If you’ve ever asked an LLM to make predictions, you’ll know it’s underwhelming. They lack specificity, play it safe, or predict things already happening. Example here.

But humans aren’t much better. We’re overconfident, biased, and bad at complexity. “That will certainly happen!” we say, before being miles off in the opposite direction.

You’ve probably experienced this yourself lately, especially with AI. If you haven’t, go make money on prediction markets. You’re clearly a prophet.

As for me, I’m a mere mortal and a software dev in the thick of it. So here are my predictions for 2026 across AI, tech and startups, to humour myself this time next year.

The Weird Stuff

  • I will forget or stop caring about this blog come end of the year.
  • We’ll see several morally questionable products emerge. Black Mirror-like.
    • My darkest guess: a company selling access to an AI child they parent as it grows up. Specifically targeting couples who can’t have kids.
    • Another: an AI dating app that promises to predict relationship outcomes, including instability and cheating. Causes people to break up preemptively.
  • A movie or TV show on a major streaming platform will heavily use AI for content (not just images or intros) and receive huge backlash. The platform will respond, but won’t condemn AI use. They have plans for more.
  • One of the financial subreddits will be taken over by coordinated AI bots. Think WSB/GameStop, but malicious: false whistleblower statements, fabricated insider knowledge, something to spark a frenzy.
  • A CEO admits their quarterly earnings call was AI-generated and not 100% accurate. Stock goes up anyway.
  • A jury will be exposed for having used LLMs, leading to an overturned verdict.
  • Similarly, a judge will be caught using LLMs through a pattern in their rulings they didn’t notice themselves.
  • An AI-orchestrated cult or religion will form, attracting extreme views and actions. (This is already just Twitter, but something more organised.)
  • AI-first dating apps will launch to replace Hinge/Tinder/Bumble and die quickly. Too early, too AI-focused in the interface. The concept will return in later years with better UX.

Big Tech

  • Google or Amazon will acquire Anthropic. Amazon already has a massive stake; Google has the existential fear of falling behind. Gemini 3 proved Google can innovate, but Anthropic are still kings of enterprise and developers.
    • If not the above, Amazon enters the AI race by purchasing a lab outright or diversifying further outside of Anthropic investment.
  • Apple will again sit this year out, waiting for the landscape to settle. They’ll make smaller, seemingly odd acquisitions that pay dividends in a few years.
  • Microsoft will try to win back developers through force of ownership: more acquisitions of beloved tools. Everyone says Cursor, but I think the inverse. Cursor stays independent, even with Microsoft circling.
  • Linear will be acquired by big tech, potentially a coding focused lab. As part of my software development expectations below, I think there will be a shift towards AI and humans working together in an interface. Linear is the best existing solution for humans and are already doing some cool stuff for agents.

Software

  • We’ll see new interfaces positioning developers as “task orchestrators” (foreshadowing an experiment…) and “reviewers” rather than “writers.” If you’ve used agentic coding, you know the feeling: waiting minutes for an agent to finish, unsure whether to context-switch or just stare.
    • Worktrees and multi-window setups help, but they’re not natural. What’s missing is proper tooling for planning, task orchestration, and code review at the pace agents demand.
    • Code is getting cheaper to produce, which raises the speed-quality bar. Iteration is fast now. The bottleneck is the interface through which developers observe, verify, and course-correct.
  • Someone will acquire Zed in a Bun-style deal, removing the pressure for immediate profitability and letting them build. Likely to be used as a harness for AI tooling. They have incredible fundamentals and aspirations, but are far off on the AI side vs other IDEs. I’d love it to be Anthropic, but I think their vision is full autonomy, not IDEs. Alternatively: Cursor makes a play to try to ditch the Microsoft-owned VSCode, or OpenAI re-enters the IDE space. Hopefully anyone but Google, due to their history killing products.
  • Many “vibe coding” platforms (the drag-and-drop, prompt-to-app tools) will hard pivot or get acquired as clear winners emerge. I expect OpenAI to be one buyer. They love consumer AI. Meta could be another; they’re behind and need a win somewhere. Meta acquired Manus since drafting this.

Wrapping Up

If there’s a thread through the weird stuff, it’s deception and malice. Manipulation dressed up as products. Bots pretending to be retail traders. Earnings calls that aren’t real. Judges and juries trusting tools they don’t understand. The line between authentic and synthetic gets blurrier, and most people won’t notice until something breaks.

The tech and software predictions: companies fight for control. Money moves, tools consolidate, developers become reviewers. The usual.

Anyway. If prediction #1 turns out to be wrong and I actually remember this blog exists, I’ll revisit in December and see how far off I was. I’d guess extremely far off, but…

See you then. Maybe.