As we enter 2026, podcast hosts Dan and Brandon, COO of Every, reflect on the remarkable growth of 2025 and predict the new trends that will disrupt the software industry this year. They foresee the rise of 'agent-native' architecture, designers gaining coding capabilities, and the transformation of traditional development approaches as key themes. They also provide a realistic analysis of the conditions needed for truly autonomous AI agents and the timeline for AGI (Artificial General Intelligence).

1. 2025 Retrospective and Explosive Growth

Before diving into predictions, the two discuss how special 2025 was for Every. Revenue had been nearly flat for 4-5 years, but starting in June 2024, the business doubled and entered an explosive growth trajectory.

A major factor behind this growth was the aggressive adoption of AI. They built and used their own tools internally (AI email assistant Kora, ghostwriter Spiral, etc.), which became a competitive product advantage. Events like 'Claude Code Camp' where they engaged with subscribers, and appearances on Lenny's Podcast that shifted public perception, were particularly effective.

"2025 was truly our year. If you look at our revenue chart, it's really interesting — the graph that was nearly flat for the past 5 years completely inflected upward this year. It was like igniting the engine. 2026 will be the year we shift from 1st gear to 5th."

They proved by example what it means to be an AI-first business. With just about 20 team members, they're doing the work that previously required a 100-person company. Thanks to AI, scaling is now possible without endlessly adding headcount.

2. 2026 Key Prediction: Agent-Native Architecture

Dan identified 'Agent-Native Architecture' as the biggest change in software for 2026. This isn't just about bolting a chatbot onto an app — it means a fundamental transformation in how software itself is designed.

He explained this change in three stages:

1. Replacing user actions: AI agents can perform every click and action that users can do in the app.
2. Executing internal code: Agents can execute code that exists within the app but isn't accessible to users through the UI (e.g., regenerating all data summaries).
3. Replacing developer functions: Agents directly modify code and add features, creating real-time customized apps tailored to individual users.

"By late 2025, we reached a point where code written by coding agents mostly works well. This fundamentally changes software engineering. I want to call this 'Agent-Native Architecture.'

Stage 1 is 'agents can do everything users can do.' But at Stage 3, 'agents can do everything developers can do.' Meaning, when a user says 'change my software this way,' the agent modifies the code and creates a custom version just for that person."

Currently, Anthropic is leading this space, and Notion is also moving in this direction.

3. The Rise of Designers and the Deepfake Threat

Designers Become Superheroes of the AI Era

As AI lowers the barrier to coding, designers who had great aesthetic sense and vision but lacked coding skills are poised to benefit enormously. They no longer need to convince developers — they can now build anything they imagine into high-quality products on their own.

"Designers are the next superheroes of the AI era. They've long had exceptional taste and vision, but had to depend on developers. Now code is cheap. Creative designers like Lucas will create incredible experiences through 'Vibe Coding.'"

Deepfakes and the Confusion of Reality Perception (Doomer Take)

Brandon also offered a darker outlook. AI-generated fake videos are becoming so sophisticated that ordinary people (especially the parent generation) can no longer distinguish reality from fabrication. Deepfakes could cause major social disruption during upcoming elections and political events, and he predicted that strong regulations or mandatory 'AI-generated' labeling will be introduced.

"My mother sent me a video of a gorilla caring for a child, and it looked so real that when I said 'Mom, this isn't real — AI made this,' she didn't believe me.

By the 2026 midterm elections, really scary things will happen. Videos that look completely real, making it appear that someone did something they never did, will flood the internet."

4. The Birth of a New Engineer: The 'Agent Engineer'

Major tectonic shifts are also expected in software engineering. Dan classified engineers into three types:

1. Traditional engineers: Those who don't use AI at all (a shrinking group).
2. AI-augmented engineers: Those who code traditionally but use AI as a tool to speed things up. They still read and write code directly.
3. Compound/Agentic engineers: These are the core of 2026. They've given up writing code directly and instead transitioned to the role of managing and directing AI agents.

These 'agent engineers' differ from mere 'vibe coders' who don't understand code. They're professionals who deeply understand software architecture but delegate actual implementation entirely to AI, focusing on higher-level design and management.

"Right now we're witnessing a schism in software engineering. There's a third group that many companies are missing — the 'agent engineers.'

These people have reinvented the craft of software engineering for the agent world. They've given up the joy of writing code line by line, choosing instead to direct multiple agents simultaneously and build much larger systems."

5. AGI and the Future of Autonomy

Finally, the two discussed AGI (Artificial General Intelligence) and AI autonomy. While 2025 was indeed the 'year of agents,' AGI prediction timelines from leaders like Sam Altman and Dario Amodei have been quietly pushed back.

Dan pointed out that while AI's pace of advancement remains fast, achieving full autonomy is a far more complex problem than initially thought.

"By the end of 2026, we'll still be saying 'wow, AI has really advanced,' but AGI won't arrive by 2027. AGI is a far more complex and difficult problem than we thought.

True AGI means a model that lives continuously inside a computer without needing to be shut off, modifying its own goals and constantly learning. Current models can work autonomously for about 20 minutes, but that's far from running indefinitely on their own."

In 2026, the focus will be on training AI to work more autonomously — for longer periods and more independently without human intervention. This suggests a direction toward giving AI a kind of 'personality' and 'agency' beyond simple command execution.

Closing

2026 will be the year AI technology goes beyond being merely a 'cool tool' to completely transforming how software is built, who builds it, and what the results look like.

• Apps will code themselves to change as users desire.
• Designers will gain the wings of coding to become the ultimate product creators.
• Engineers will no longer be people who write code, but conductors who direct armies of AI.

Even though AGI won't arrive as soon as next year, the way we work and use software will undergo fundamentally irreversible change over the course of 2026.