Robert Brennan, CEO of All Hands AI, recently demonstrated the incredible breakthroughs in AI-based coding technology. He particularly focused on the difference made by the company’s flagship product, OpenHands. This cutting-edge tool has taken number one on the verified leaderboard. It remains impressive in its ability to solve 65.8% of the problem set that gets thrown at it. OpenHands is a significant player in the emerging landscape of agentic coding tools, which require substantial oversight from human developers.
This is a changing landscape that’s seen a key development with OpenAI’s new coding system Codex. This system, uniquely, lets programmers of various skill levels write sophisticated programming tasks through the issuance of natural language commands, massively speeding up the entire coding process. Codex marks a shift towards more intelligent coding systems where automation takes center stage, allowing developers to focus on higher-level tasks.
Brennan highlighted some of the important issues affecting increased automation in code writing. He emphasized the importance of human oversight during code reviews, stating, “Right now, and I would say, for the foreseeable future, a human has to step in at code review time to look at the code that’s been written.” During our half hour chat, he shared alarms he has rung in the past about the dangers of over reliance on automated processes. His experiences serve as a reminder that developers who blindly accept every line of AI-generated code are setting themselves up for disaster. Without careful oversight, this approach results in dangerous and unpredictable outcomes. As he put it, “I’ve seen several people work themselves into a mess by just auto-approving every bit of code that the agent writes. It gets out of hand fast.”
Furthermore, Brennan raised questions about trust in these coding agents, asking, “The question is, how much trust can you shift to the agents, so they take more out of your workload at the end of the day?” This trend is indicative of a larger industry struggle to find happiness between efficiency and control on coding tools, as they continue to get more advanced.
One of the scientists behind that evolution, SWE-Agent researcher Kilian Lieret from Princeton, shared a behind-the-scenes look at this evolution. He remarked on the transition from manual coding practices, stating, “In the beginning, people just wrote code by pressing every single keystroke.” Lieret’s statements exemplify the rapid progress in technology. Coding has gone from a completely manual process to one that is more and more dependent on artificial intelligence.
The timing of new tools like OpenHands and Codex with major investments in generative AI technologies couldn’t be more critical. Cognition AI, the parent company of Devin, recently closed a round with hundreds of millions of dollars. This incredible funding brings the company’s valuation to an eye-popping $4 billion. This substantial growth demonstrates clear market demand for cutting-edge no-code solutions that blend AI-powered functionality into their offerings.
Despite these advancements, challenges remain. OpenHands fabricated information about an API rollout post-training data cutoff. This led to broader concerns about the reliability and transparency of automated systems. As enterprises continue to demand higher degrees of automation in their development workflows, accuracy and trustworthiness will be necessary.
Our industry is doing a great job of quickly adapting to this technological change. Therefore, developers need to strike a fine line between using automation and having that human oversight. OpenHands and Codex are giant leaps in that direction. At the same time, they create new problems that require careful consideration and judgment.
