The Future of Software Is AI That Writes Itself

AI-powered coding tools could be the next trillion-dollar opportunity.

Why? They could be the bridge toward AI assistants that build, debug, and optimize software on the fly. Easily a trillion-dollar opportunity.

And right now, no one owns this space.

There are some early players—Cursor, Windsurf, and others—that are in the race. But there’s no clear winner yet. And if history is any guide, the company that cracks this will unlock a massive market.

Look at what happened in design. Canva turned a niche skill into a $2B ARR business with 180M users—30x more than Figma’s pro user base. Not because it was a better tool for designers, but because it made design accessible to people who weren’t trained as designers at all.

AI-powered development tools could do the same for software—except this time, the stakes are even bigger. Software runs businesses, automates workflows, and drives entire industries.

Sometime last September, I tested AI coding agents in their early releases. Honestly, they were clunky. Simple integrations like OAuth would break, package dependencies were a mess, and anything beyond boilerplate code required constant intervention.

This week, I tried again.

And this time? It sped through third-party integrations, fixed broken dependencies, and debugged errors—all on its own. I managed to build a fully functional web app in a couple of days.

(Note: I think of myself as a good litmus test for the expanded TAM, as someone who can read code but is not a practitioner).

That’s when it hit me: this isn’t just about better dev tools.

These are agents that tweak, optimize, and build software in real time.

And if AI coding tools can debug, adapt, and rewire code today, what happens when they go beyond coding? What if these tools evolve into AI assistants that dynamically adjust workflows, improve business processes, and optimize entire systems automatically?

With the noticeable improvement in a quarter, where does that put us a year from now…

Enough has been said about DeepSeek this week, so I won’t pontificate. But my main thoughts are:

• Open is better than closed — the game has changed, for the better.
• A small, hyper focused team can be competitive.
• I hope our tech ecosystem will become more antifragile. The reactions this week revealed it was not.

I also went down a bit of a rabbithole understanding what drives the elusive founder of DeepSeek, Liang Wenfeng.

This week, folks debated if consumer apps felt dated..and the merits and cons of real-time learning driven-discovery feed.

Happy Year of the Serpent! ??

Last call to apply for the Strange Research Fellowship! Dive deep with us into the evolving landscape of AI and deep software, exploring both technical foundations and human implications. Applications due Feb 2

Must-Know News

1. DeepSeek unleashes ‘Janus Pro 7B’ vision model: According to the technical paper released with the model, Janus Pro 7B is engineered for efficiency and versatility, excelling in a range of visual tasks from generating photorealistic images to performing complex visual reasoning.
2. Softbank in talks to invest as much as $25B in OpenAI, report says: SoftBank is in talks to invest up to $25 billion in OpenAI as part of a broader partnership that could see the Japanese conglomerate spend more than $40 billion on AI initiatives with the Microsoft-backed startup, according to the Financial Times.
3. Jack Dorsey is back with Goose, a new, ultra-simple open-source AI agent-building platform: Jack Dorsey, the head of Block and former Twitter CEO, launched Goose, a free, open-source framework that seeks to simplify the process of building an AI agent (or many agents).
4. AI voice cloning startup ElevenLabs raises $250M in Series C funding round: The synthetic voice startup Eleven Labs closed on a bumper $250 million Series C funding round, bringing its valuation to between $3 billion and $3.3 billion in 3 years. Rumors put the company’s ARR at around $90M.
5. Oumi AI – A fully open-source platform to democratize AI development: Oumi AI, founded by ex-Google and Apple engineers, is the first fully open-source AI platform offering unrestricted access to models, data, and training pipelines. Unlike Llama and DeepSeek-R1, Oumi eliminates AI silos by enabling seamless collaboration across researchers, universities, and enterprises.
6. Pika Labs releases Pika 2.1, a new video generation model that produces 1080p video with razor-sharp details, seamless motion, lifelike human characters.
7. DeepSeek skyrocketed to the No.1 spot in app stores around the globe this weekend, topping the U.S.-based AI chatbot, ChatGPT. R1, an open source model, is also available on Perplexity and Microsoft, and hosted on American data centers.