General Artificial Intelligence (AGI) is no longer a distant future but has become a reality with the emergence of "long-horizon agents." According to an article titled "2026: This is AGI" released on the 14th by Sequoia Capital partners Pat Grady and Sonya Huang, while technical definitions of AGI remain debated, from a functional standpoint, AI capable of autonomously solving problems has officially arrived, and 2026 will be their year.
Sequoia Capital stated that coding agents are the first concrete instance of AGI deployment, with more types of agents rapidly emerging. Unlike earlier conversational AIs, the new generation of long-horizon agents can reason based on baseline knowledge and achieve goals through continuous self-iteration, much like humans. This leap in capability marks a transformation for AI from a mere "conversationalist" to a practical "executor" that can deliver work.
This shift will have a profound impact on the business and investment landscape. Sequoia's analysis suggests that as agent capabilities grow exponentially, the fundamental logic for founders building products will change—from selling software to directly "selling work outcomes." Future AI applications will no longer be just auxiliary tools but entities that can work alongside humans as "colleagues" around the clock, with users transitioning from independent contributors to managers of agent teams.
With recent breakthroughs like Claude Code and other coding agents crossing critical capability thresholds, the market's perception of AGI has been reshaped. The article emphasizes that through enhancements in reinforcement learning and agent architecture optimization, agents' ability to handle complex tasks is doubling approximately every seven months, which will fundamentally alter corporate talent structures and productivity boundaries.
Sequoia Capital indicated that as investors, they prefer to avoid technical definition debates around AGI and instead propose a pragmatic functional definition: AGI is the "ability to solve problems on its own." For businesses focused on outcomes, how the AI achieves the goal is less important than its ability to actually complete the task.
The article breaks down AI possessing this capability into three core components:
Baseline Knowledge (Pre-training): This was the core driver of the 2022 ChatGPT moment. Reasoning Capability (Inference-time Computation): Achieved with the release of models like o1 by the end of 2024. Iteration Capability (Long-horizon Agents): This is the latest breakthrough, where AI can work autonomously for hours, correct errors, and decide next steps without specific instructions, akin to a generally intelligent human.
To illustrate what "solving problems on its own" means, the article uses a recruitment scenario: when a founder needs to find a Developer Relations lead who is both technical and active on social media, the traditional approach is to post a job description. An agent, however, can autonomously execute a complex search loop.
According to the article's description, an agent can complete the mental cycle of a human recruitment expert in just 31 minutes: it would search LinkedIn for relevant roles at competitor companies like Datadog and Temporal, then switch to YouTube to filter for speakers with high engagement, and further cross-reference activity levels and content quality on Twitter. The agent can even detect potential attrition signals by analyzing decreases in posting frequency, ultimately identifying the best candidate and drafting a personalized outreach email.
This ability to operate in ambiguous environments—forming hypotheses, testing, making mistakes, and adjusting course until the goal is achieved—is the core characteristic of long-horizon agents. Although they currently still produce hallucinations or lose direction, their development trajectory is irreversible, and errors are becoming increasingly correctable.
Regarding how this leap was achieved, Sequoia Capital points out that enabling models to "think" for extended periods is challenging. Two technical paths have proven effective and scalable: One is Reinforcement Learning, primarily led by research labs. This involves continuously "nudging" and guiding the model during training to teach it to maintain focus over long durations. Significant progress has been made in multi-agent systems and tool-use reliability. The other is Agent Harnesses, which fall into the application layer. Developers design specific scaffolds (e.g., memory handoffs, compression) to work around known model limitations. Highly-regarded products on the market, such as Manus, Claude Code, and Factory's Droids, benefit greatly from superior architectural design.
Tracking by METR of AI's ability to complete long-horizon tasks shows exponential progress in this field. Projecting the current trend, agents will be able to reliably complete work that takes a human expert a full day by 2028, and a full year's worth of work by 2034.
"Can you hire an agent?" Sequoia Capital believes this is the litmus test for AGI. The current market landscape shows that specialized agents are rapidly emerging across various sectors, from OpenEvidence in medicine and Harvey in law to XBOW in cybersecurity.
This signifies a massive paradigm shift for entrepreneurs. AI applications in 2023 and 2024 were mostly "conversationalists" with limited impact; applications from 2026 onwards will be "executors." This shift makes "selling work" possible. Founders need to rethink: under this new paradigm, which ongoing, attention-requiring tasks can be handed over to agents? How should pricing and packaging be oriented towards "outcomes" rather than "tools"?
The article concludes by urging the market to "Saddle Up" and prepare for the exponential growth of long-horizon agents. While today's agents might only work reliably for about 30 minutes, they will soon handle a full day's workload, and eventually even tasks equivalent to a century of human work.
This means that roadmaps once considered too ambitious—such as cross-referencing 200,000 clinical trial data points or refactoring the entire US tax code—are now becoming feasible. In the dawn of the AGI era, ambitious plans are steadily transforming into realistic business strategies.
Comments