What you can actually build without code in 2026: the software map

The question I get from non-technical friends is never “can AI build software?” anymore. It’s “what can I build?” Fair question, and the answer is more specific than the hype suggests. Some software is a weekend project now. Some is still a trap. Here’s the map.

A quick note on terms, since I’ll use them throughout: a web app is software that runs in a browser (think Google Docs). A mobile app is installed from an app store. SaaS (“software as a service”) isn’t a type of software at all, it’s a business model: software you charge a subscription for. An automation is a small program with no interface that just does a task on a schedule or a trigger.

Web apps: the default starting point

If you’re building your first thing, build a web app. It runs in any browser, there’s no app store review, and updates go live the second you publish. AI agents are at their best here because the whole stack (the pages you see, the server logic, the database) is mature, well-documented territory that models have seen a million times.

Realistic weekend projects: a dashboard that tracks something you care about, an internal tool for your team, a directory or listing site, a booking page, a small tool people use once and bookmark. This site is a web app. So is every product I’ve shipped in the last year.

Mobile apps: doable, with friction

Agents write perfectly good iOS and Android code. The code isn’t the hard part anymore. The friction is everything around it: a $99/year Apple developer account, app review that can take days and reject you for reasons that read like riddles, and screenshots, privacy labels, and age ratings before anyone can download anything.

My advice: unless the idea genuinely needs to live on a phone (notifications, camera, offline use), start with a web app and see if anyone uses it. A web app that people love can become a mobile app later. The reverse path is painful.

SaaS: the software is the easy part

Building the software side of a SaaS is very achievable now. Agents wire up login, payments through Stripe, and per-user data without drama. What makes SaaS hard has nothing to do with code: you’re asking strangers for money every month, which means pricing, churn, support emails, and tax compliance.

If your real goal is “make money from software,” know that a $5 one-time purchase on a marketplace is a much shorter path than a $20/month subscription. SaaS is a great second business. It’s a rough first one.

Automations: the underrated first win

The fastest real win for a non-engineer isn’t an app at all. It’s an automation: a script that renames and files your invoices, a bot that posts a daily summary to your team’s chat, a job that checks a competitor’s pricing page every morning and emails you when it changes.

No interface to design, nothing to deploy for the public, and you learn the core loop of working with an agent (describe, review, run, adjust) in an afternoon. Everyone skips this category because it’s unglamorous. Don’t.

What to skip for now

Honesty section. These are still hard even with agents:

Real-time multiplayer anything. Games, live collaboration, chat at scale. The code isn’t the problem; the infrastructure judgment is.
Hardware and firmware. Agents can’t hold the soldering iron, and the feedback loops are slow.
Regulated territory. Health records, money movement, anything with compliance obligations. You can prototype, but shipping without professional review is asking for it.
Anything you wouldn’t know how to test. If you can’t tell whether the output is correct, you can’t safely ship it. That rule will outlive every model release.

How to pick your first build

Pick the smallest thing you would actually use yourself. Not the thing that sounds like a business. The ladder that works: an automation first (afternoon), then a web app (weekend), then, only if people use it, a mobile app or a paid product. Each rung teaches you exactly one new skill, and agents flatten each one a little more every few months.

The old promise of no-code was “anyone can build software,” and it was mostly false. The 2026 version is narrower and true: anyone willing to learn how to direct software can build it. The directing is a skill. It’s also a much shorter curriculum than programming ever was.

If you’re picking your first build and want a second opinion on scope, email me: contact@softwarewithoutcode.com.