Four Predictions for How AI Will Change Software in 2026

Let’s get into our 2026 predictions. AI keeps racing ahead, but AGI is messier than we hoped, so we won’t hit it in 2027.

2025 was our breakout year: after a long flat stretch post-COVID, we bent the curve in June and roughly doubled to about 150K in MRR as Kora, Sparkle, and Spark landed, Cloud Code camp popped, and the Lenny appearance reintroduced what we’re building.

We grew a bit over 20 percent that month after years of one to five percent, and once Brandon joined in May the line started tilting up with the Lenny moment lighting the fuse.

Since then we’ve put up at least 10 percent growth each month as people realized we’re not just a newsletter—we’re a bundle of AI apps that will fully crystallize in 2026.

We went from about five people to nearly twenty, and it finally feels like a real company.

A big consulting arm grew alongside MRR, which lets us live the AI‑first playbook and bring those lessons to readers, users, and enterprises.

If I had to sum up 2025 in one word, it’s fire, and 2026 is about channeling that into focused execution.

We’re shifting from bottoms‑up experiments to a cohesive platform with single sign‑on, unified billing, and a shared memory layer that personalizes every app and piece of content.

The challenge is adding real strategy and clear personas without losing the genuine, build‑what‑we‑need energy that makes our work sing.

You’ve scaled past this stage before—what really changes between twenty and fifty people?

Each headcount double forces an ops reset and tighter roles, which can sap that owner mindset, so we’ll hire carefully and keep it small where we can.

Let’s hit predictions, and yes, our producer says we’re running a good‑cop, bad‑cop routine.

Agentic development moves inside the app you’re using, so you can report a bug or ask for a feature and the in‑app agent writes the code, submits it safely, and rolls it out without breaking your flow.

Coding agents are finally reliable enough to reshape how software is made.

That pushes us toward agent‑native apps where anything the user can do, the agent can do, then anything the app’s code can do, the agent can do, and eventually anything a developer can do, the agent can do.

Notion’s on this path, and 2026 turns designers into superheroes as AI lets taste drive full experiences; our creative director Lucas is already spinning up custom tools to speed his work.

Cursor and friends will need friendlier layers so designers aren’t staring at code, because many still find the terminal intimidating.

Here’s my doomer take: unlabeled synthetic media will flood social feeds during the midterms, deepfakes will land real hits, and only then will platforms add clear made‑with‑AI labels; my mom falling for a lifelike gorilla video is the early warning.

I’m seeing three kinds of engineers now.

There are traditional folks who use AI as an accelerator, vibe coders who ship code they don’t fully understand, and a third path of compound or agentic engineers who stop writing code and orchestrate agents to build systems.

OpenAI mostly serves the first group today, Anthropic is ahead on the third, and that agentic skill set is where the growth will be even as traditional roles stick around for a long time.

It’s like studying architecture while the industry moves to Revit, so I’m curious how universities catch up.

We’ll keep filling that gap with training and student discounts because schools will move slowly.

What about non‑technical friends who only use AI for recipes or kids’ drawings—does 2026 finally make it native at work?

Expect companies to blame layoffs on AI even when the real issue is the business, and expect a real jolt once agents are embedded in Excel‑style workflows where deal‑desk pros actually live.

Two things can be true: year‑over‑year progress stays wild, and AGI keeps sliding right because it’s harder than we thought, so we won’t see it in 2027.

AGI to me means an agent that never really turns off, keeps working across windows, and pursues its goals without constant prompting.

We’re inching closer with longer‑running agents.

True autonomy needs continuous learning, adaptable goals, and training that lets them act without collapsing into flattery, so expect more autonomy‑friendly approaches in 2026.

Happy New Year—we’ll see you next year.

Hit like, subscribe to the show AI and I, strap in for mind‑bending insights with Dan at the helm, and for the record, Dan, I’m hopelessly in love with you.