#ai
**Source:** [Sequoia Capital - 2026: This is AGI](https://sequoiacap.com/article/2026-this-is-agi/)
## Functional Definition of AGI
Rather than pursuing abstract technical definitions, Sequoia defines AGI pragmatically - **the ability to figure things out**. This requires three components working in concert:
1. **Baseline Knowledge** (pre-training) - Large models trained on vast data
2. **Reasoning Ability** (inference-time compute) - Deep thinking capability (e.g., o1 reasoning models)
3. **Iterative Problem-Solving** (long-horizon agents) - Autonomous iteration over extended timeframes
All three are necessary. Knowledge without reasoning is pattern matching. Reasoning without agency is still just a chatbot. Agency without knowledge fails immediately. The convergence of all three marks the moment we've crossed into functional AGI.
## The Three Technologies That Converged
- **ChatGPT (2022)** provided the knowledge layer
- **Reasoning models like o1 (late 2024)** added sophisticated thinking at inference time
- **Recent coding and task agents** added the capacity for autonomous, multi-step iteration
## Shift from "Talkers" to "Doers"
The substantive change is moving from interactive tools (ask a question, get an answer) to systems that work **autonomously for hours or days**. You can now run multiple agent instances in parallel on different subtasks.
This reframes the productivity bottleneck: it's no longer "per-interaction speed" but rather "how many things can I delegate continuously?"
## Exponential Scaling Trajectory
METR's research shows agent capability roughly doubles every 7 months:
- **By 2028:** agents handling full-day expert tasks (what a specialist completes in 8 hours)
- **By 2034:** agents capable of year-long projects
This isn't linear improvement—it's exponential capability growth.
## The Strategic Question for Business
**What's the work requiring sustained, persistent attention where the main bottleneck is "nobody can afford to have someone working on it continuously"?**
That's where agents create immediate value. Not in replacing experts, but in tackling problems that were economically infeasible before because continuous human effort was required.
Examples: specialized medicine, law, cybersecurity, recruiting, design work.