Stanford course and Monaco engineering team adapt to coding agents in software workflows
CS146S revises assignments after students one-shot prior homework with agents; Monaco builds internal workspaces to operationalize coding agents for engineering and sales.
1 source · cross-referenced
- Stanford’s CS146S revised coursework after students used coding agents to complete prior-year homework in a single attempt.
- Monaco’s engineering team adopted coding agents and built internal workspaces to run, test, and deploy patches.
- Course assignments now require students to contribute to open-source projects and have their patches reviewed by maintainers.
- Monaco uses Devin for code review and built custom agent workspaces to handle background tasks like reproducing failures and testing patches.
Stanford’s CS146S course, taught by Mihail Eric, recently revised its assignments after observing that students could one-shot prior-year homework using coding agents. In one exercise, students previously ran an AI code reviewer on a repository and evaluated its feedback; the course team found that students could now complete such exercises in a single attempt. As a result, the current iteration of the course requires students to contribute to an open-source project and have their patches reviewed by the repository maintainer. According to Eric, this change preserves the learning value of understanding maintainer intent, code behavior, and potential side effects—skills that cannot be fully automated by current agents.
At Monaco, Mihail Eric leads the AI engineering team, which has operationalized coding agents in its software workflow. The team evaluated third-party code review tools and ultimately continued using Devin for that function. However, for background tasks such as reproducing failures and testing patches, Monaco built its own agent workspaces because the off-the-shelf products did not fit the company’s cloud setup. Eric describes the approach as: “You buy to learn, and you build to keep.” The internal system, referred to as Hal, enables engineers to ask the system to investigate incidents and open pull requests, while sales teams can query the same system for product explanations tailored to customer conversations.
The course and company illustrate a broader shift: educators and engineering organizations are redesigning assignments and infrastructure to account for agents that can write, review, and test code. In CS146S, students now confront the human elements of maintainer review and intent that agents cannot resolve. At Monaco, the engineering team has built custom workspaces to ensure agents operate within the company’s specific cloud and permission constraints, reflecting the operational realities of deploying agentic systems at scale.
- Nov 11, 2026 · Latent Space — swyx
OpenAI demonstrates voice steering, cloud handoffs, and personal agents for Codex at AI Engineer Code Summit
Trust79 - Nov 11, 2026 · The Verge — AI
OpenAI to launch verified AI Business Agents in ChatGPT for US e-commerce and service businesses
Trust79 - Nov 10, 2026 · Latent Space — swyx
NaiveAI releases Naive-N0.5-Flash coding model with ultrafast token throughput
Trust74