With AI trivializing the mechanical act of writing code, the most valuable traits for emerging engineers are no longer just technical proficiency. Instead, employers will seek demonstrated agency (the drive to build), taste (knowing what to build), and a commitment to quality.
Automating coding tasks won't eliminate engineers. Similar to the shift from assembly to higher-level languages, AI tools increase output potential, leading to an explosion in demand for software and the builders who can leverage these powerful new platforms.
The PM role is intentionally undefined, meant to adapt to a team's needs—from strategy to quality control. However, these functions can often be filled by a strong engineering lead or designer, making dedicated PMs non-essential, and potentially harmful, on smaller teams.
The dominant AI interface will be a universal conversational layer (chat/voice) for any task. This will be supplemented by specialized graphical UIs for power users needing deep functional control, much like an executive sometimes needs to edit a document directly instead of dictating to an assistant.
OpenAI's Head of Codex argues the main barrier to AGI isn't model capability but human laziness and lack of creativity in prompting. People use AI tens of times daily, but the potential is for tens of thousands. The friction of typing and thinking of prompts is the key limiter.
The threat of AI to SaaS is overstated for companies that own either a deep relationship with the user or a critical system of record. "Glue layer" SaaS companies without these moats are most at risk, while those like Salesforce (owning the customer relationship) are more durable.
The manual management of deployment and monitoring will become obsolete. A new, fully AI-managed stack will emerge, allowing founders to simply ask an agent to build and iterate on products. The company's main communication tool may even become the interface for managing these agents.
OpenAI believes it has sufficient coding data. The next data advantage lies in capturing "knowledge work" tasks—data not on the public internet. This may require novel approaches like acquiring failed startups for their internal data from tools like Slack.
In an unusual strategy, OpenAI provides its latest models to direct competitors. The company believes that a more competitive market accelerates learning and pushes them to improve faster. This long-term view prioritizes the overall distribution of intelligence over short-term competitive moats.
A lesson from Dropbox's competition with Slack is that users gravitate towards a "center of gravity" or system of engagement, even if it's less optimal. AI tools that become the primary, easy-to-use interface for work will win over those built solely as backend workflow automation.
Instead of relying solely on top-down, consultant-led workflow automation, enterprises should empower individual employees with AI tools. This builds user fluency and intuition, allowing them to pull AI into their own workflows, resulting in greater overall impact and less disempowerment.
OpenAI's strategy for agents is a three-step journey: 1) Perfect agents for software engineering. 2) Provide open-ended tools for tinkerers to discover general use cases. 3) Use learnings from tinkerers to build highly productized, specific features for the mass market.
The focus of "code review" is shifting from line-by-line checks to validating an AI's initial architectural plan. After plan approval, AI agents like OpenAI's Codex can effectively review their own generated code, a capability they have been explicitly trained for, making human code review obsolete.