what sign up
nixm nixm

← learn

The future of AI: the next five years, sorted into likely, possible and hype

October 14, 2026

The near future of AI is less dramatic and more pervasive than the headlines. Likely by 2030: AI agents doing multi-step work on their own, models that handle text, images, sound and video together, AI built into every office tool and most devices, much cheaper models and good ones that run on your own hardware, and real regulation. Possible: a decade of work reorganised around AI, with new jobs that do not exist yet. Hype, for now: a conscious machine, or one that replaces human judgement rather than human typing.

How to read predictions about AI

The field has a pattern: short-term predictions are too bold and long-term ones too timid. Nobody in 2019 expected a model that writes working code by 2023; most people in 2023 expected the office to look different by 2025, and it mostly does not. So: the capability arrives fast, the world absorbs it slowly. Everything below is sorted by that rule.

Likely by 2030

  • Agents, everywhere. AI that is given a goal and does the steps: research, booking, coding, paperwork. The AI agents page explains the shape. Expect them to be ordinary, supervised, and good at the boring half of office work.
  • One model, every medium. Text, images, audio and video in and out of the same system. You will talk to it, show it things, and get a picture, a clip or a document back. nixm already works this way in one conversation; the tools around it will too.
  • AI inside the tools, not beside them. The chatbot in a separate tab is a transitional form. The next step is AI in the document, the spreadsheet, the inbox, the camera, the car, where you do not open it, it is just there.
  • Much cheaper, and local. The cost of a given capability has been falling fast and will keep falling. Good models will run on a laptop or a phone with no server, which changes the privacy picture for the better for people who choose that.
  • Real rules. The EU AI Act of 2024 phases in; other jurisdictions follow with their own. Disclosure of AI-generated content, obligations for high-risk uses, and settled answers on training data are coming, unevenly.
  • The energy and data fights get louder. Data centres, water, electricity, and who owns the material the models learned from. These are the political questions of the decade.

Possible, with wide uncertainty

  • Work reorganised. Not mass unemployment, but a slow restructuring: routine knowledge work shrinks, the junior rung changes, new roles appear around directing and checking AI. The jobs page has the detail.
  • Science at a different pace. AI-driven discovery in biology, materials and medicine is the candidate for the decade's biggest payoff, and the least visible to a normal person.
  • AGI, by some definition. Whether the next few years bring something fairly called general intelligence depends mostly on which definition you use; the AGI page explains why the dates disagree.
  • A plateau. Equally possible: current methods level off, and the next five years are about deployment rather than breakthroughs. Plenty of researchers expect this.

Hype, for now

A conscious machine. An AI that wants things. Robots doing a plumber's day. A system that replaces human judgement and responsibility rather than human typing. These may come eventually; nothing in current systems points to them within five years, and planning around them distracts from the changes that are already here.

What to prepare for, personally

  1. Learn to direct and check AI, because that is the job that grows while producing first drafts shrinks.
  2. Keep the skills the tool makes easy to skip. Writing, arithmetic, reading a contract. The people who keep them will be the ones who can tell when the tool is wrong.
  3. Decide now what you will not hand over. Your private questions, your health, your kids' faces. The tools will get more capable and more embedded; the defaults on data will not improve unless you choose tools where they already have. nixm's are fixed: no transcript, no training on you, conversations that dissolve.

frequently asked

What is the future of AI?

Over the next five years, likely: AI agents doing multi-step work, models that handle text, images, audio and video together, AI built into everyday tools, much cheaper and local models, and real regulation. Possible: work reorganised and faster science. Hype for now: conscious machines.

Will AI take over the world?

Nothing in current systems points to that within the foreseeable future. The realistic concerns are jobs, privacy, misuse and concentration of power, which are already here.

What jobs will exist in the future because of AI?

Roles around directing, checking and integrating AI, plus the work that stays human: physical skill, accountability, trust and judgement. See the article on which jobs AI replaces.

Will AI replace humans in the future?

It replaces tasks, especially routine knowledge work, and expands what one person can do. Judgement, goals and responsibility remain human on any evidence we have.

What is the future of AI in education?

A tutor for every student that explains at any level, with the open problem that the same tool can do the assignment. Expect schools to shift assessment toward understanding and process.

How should I prepare for the future of AI?

Learn to direct and check AI, keep the skills it makes easy to skip, and decide which parts of your life you will not hand to tools that keep your data.

try it now — no account, no transcript kept. even the signal dissolves in seven days.

seed a signal →

Enter your email and receive a link to sign in. No passwords, just a secure signal to connect.

why connect

  • 20 free credits to start
  • your signals, remembered
  • buy more credits anytime
  • image & video generation features
or
+