Key Moments
Opencode CEO: Blocked, 20X Growth in 6 Months, Building the Coding Agent for the World
Y CombinatorWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
Coding agent growth exploded after Anthropic blocked Opencode, leading to 13M users and $40M annual revenue in 8 months, challenging closed-source models.
Key Insights
Opencode achieved 13 million monthly active users and 20x growth in 2024, processing 7 trillion tokens daily, surpassing OpenRouter's total.
Anthropic's move to block Opencode by rejecting requests mentioning the name, inadvertently boosted Opencode's visibility and user acquisition.
Open-source models, once lagging 6 months behind frontier models, are now competitive for real work, driving significant adoption through platforms like Opencode.
Opencode processes more tokens from open-source models than all frontier models combined, with DeepSeek Flash, DeepSeek Pro, and GLM 5.2 being top choices.
Opencode sees substantial global usage, with China at 17% and developing countries like Indonesia (4%) and Brazil (5%) showing high engagement due to cost-effectiveness.
Fortune 500 companies, including a dozed of the top companies, are using Opencode not just for cost savings but for flexibility and to avoid vendor lock-in.
Explosive growth driven by open-source adoption and user choice
Opencode has experienced a dramatic surge in growth, reaching 13 million monthly active users and achieving a 20x increase since the beginning of the year. The platform now processes approximately 7 trillion tokens per day, a figure that exceeds the total daily token processing of OpenRouter. This rapid expansion is largely attributed to the increasing viability and adoption of open-source AI models. Opencode's subscription product, launched just eight months prior, is on track to generate $38-40 million in annualized revenue, with its monthly subscriber base growing to 160,000. The company's success is amplified by its open-source ethos, allowing users to leverage any AI model, which particularly appeals to a global audience where expensive proprietary solutions are out of reach.
Anthropic's blocking tactic inadvertently fueled Opencode's rise
A pivotal moment for Opencode occurred in early January when Anthropic began attempting to block users from running Claude Code subscriptions through Opencode. The method involved rejecting requests if the system prompt simply mentioned the word 'Opencode.' While seemingly a defensive move, this action inadvertently put Opencode and Claude Code on equal footing in the eyes of many users. It generated significant buzz, drawing attention to Opencode for the first time for many who weren't previously aware of it. This became a case study in how, similar to Instacart's experience with Amazon's Whole Foods acquisition, a perceived threat can actually drive growth and adoption by increasing product visibility and user curiosity.
Global reach and cost-effectiveness in developing economies
A significant driver of Opencode's growth is its strong performance in developing countries. Countries like Indonesia (4% of traffic), Brazil (5%), and Vietnam represent a large user base where a $200 monthly cloud code subscription is prohibitively expensive. Opencode's platform provides access to powerful coding agents at a much lower cost, making advanced AI tools accessible to a broader global demographic. This global demand also highlights a shift in the market, where the 'magic moment' of experiencing a coding agent is now being democratized, moving beyond high-income markets.
Open-source models achieve parity with frontier models
Initially, Opencode primarily facilitated the use of proprietary model subscriptions. However, by August-September of the previous year, the emergence of capable open-source models, such as GLM, Kimi, and MiniMax, marked a turning point. These models, which initially lagged about six months behind frontier models, rapidly improved to the point where they became viable for real-world development tasks. This convergence significantly boosted Opencode's utility, allowing users to access powerful, cost-effective open-source alternatives and driving a wave of new users to the platform.
The Gemini 2.5 shift and the rise of subscription models
A notable data point occurred in February of this year when, for a four-week period, Opencode observed users favoring Gemini 2.5 over Anthropic's Sonnet and Opus models combined. This unprecedented shift in user preference indicated that open-source models were not only competitive but, in this instance, even preferred for substantial workloads. This trend validated the hypothesis that these models were now suitable for 'real work,' paving the way for Opencode to confidently launch its subscription product, recognizing the market's readiness to pay for reliable access to these advanced, yet more affordable, models.
Diverse usage patterns: token budgeting and specialized model strengths
Opencode's data reveals that while cost is a major factor, other elements also drive model choice. DeepSeek Flash is heavily utilized, partly due to its low cost, which allows users to extend their usage, especially as they approach daily or weekly limits. Speed is another critical factor; some models offer higher tokens per second, providing a near real-time experience that contrasts with the feel of some frontier models. Additionally, users perceive specific models like GLM 5.2 as being superior for tasks like front-end design, influencing their selection for particular development needs.
Enterprise adoption driven by flexibility and avoidance of vendor lock-in
Beyond cost savings, a substantial number of large US companies, including a dozen of the Fortune 500, are using Opencode. The primary driver for these enterprises is the desire for flexibility and to avoid being locked into a single model or provider. Opencode serves as a neutral platform, allowing them to experiment with and switch between models as technology evolves. Companies are reaching out to Opencode not through traditional sales channels, but by simply submitting security questionnaires when internal teams have already adopted the product, indicating strong organic adoption and product-market fit. Some enterprises also seek features for managing token spend across different departments or limiting access to specific model tiers.
A long journey of iterative development and unwavering persistence
Opencode's apparent overnight success belies a 16-year history for its legal entity, which was incorporated in 2010. The founders, Jay V and Frank, embarked on their entrepreneurial journey in university, initially exploring various ideas and applying to Y Combinator multiple times with different proposals, including a serverless platform. This decade-plus period was marked by learning, building, and pivoting, including a consumer product that yielded valuable experience in acquisition and metrics. This long, winding path, often involving living frugally and facing numerous dead ends, forged the resilience and expertise needed to capitalize on the current AI wave, positioning them to transform their accumulated knowledge into Opencode's explosive growth.
Mentioned in This Episode
●Software & Apps
OpenCode Usage Breakdown by Model (Token Volume)
Data extracted from this episode
| Model | Token Volume (per day) | Unique Users (per day) |
|---|---|---|
| DeepSeek Flash | High | 38k |
| DeepSeek Pro | High | 31k |
| GLM 5.2 | High | 30k |
OpenCode Geo Breakdown (User Traffic %)
Data extracted from this episode
| Country/Region | Percentage of Traffic |
|---|---|
| China | 17% |
| Indonesia | 4% |
| Brazil | 5% |
| Vietnam | N/A |
| USA | Growing Significantly |
Common Questions
OpenCode is an open-source alternative to AI coding agents like Cloud Code and Codex. It allows users to access and utilize various AI models, providing more flexibility and choice compared to proprietary solutions.
Topics
Mentioned in this video
More from Y Combinator
View all 607 summaries
31 minHow Two French Engineers In New York Built The Company That Monitors The Entire Cloud
23 minThe Model-Agnostic AI Platform Betting That No Single Lab Will Win
29 minWhy Ambitious Startup Ideas Are Actually Easier To Sell
75 minThe Key Thing Human Brains Have That AI Is Trying To Learn
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free