Boris Cherny, February 2026: generated code still gets automatic and human review. Prototype code is a different case.
9
If the support process is predictable, a clear workflow may be more useful than a committee of agents.
13
Part 7 of 7 - final: Use a productivity gain only to shrink, or use it to do more? Mollick's brewing analogy asks leaders to consider the bigger path. The series ends with a choice about what the team can become. #AI #FutureOfWork
11
Part 6 of 7: 'You'll work with AI' leaves the real questions unanswered. What changes in the work? Who benefits from the time saved? Mollick argues leaders need a concrete answer. Next: choose growth, not just cuts. #AI #Leadership
8
Part 5 of 7: The people who know the work can often judge whether AI's answer is useful. Mollick makes the case for starting with internal expertise, not waiting for a perfect outside playbook. Next: why a vague AI vision fails. #AI #AIAdoption
9
Part 4 of 7: A job is a bundle of different tasks. That gives teams two practical questions: what can AI take off your plate, and what can it help you do better? Next: your experts are already inside. #AI #FutureOfWork
12
Part 3 of 7: Is the analysis good enough? What is the error rate? Mollick's point: build tests around real work, not just model rankings. Next: redesign tasks, not just job titles. #AI #Workflows #AIAdoption
7
Part 2 of 7: People may hide AI use when better results look like a threat to their jobs. Mollick argues that trust and incentives are part of the deployment problem. Next: build a test for your actual work. #AI #Workflows
7
Part 1 of 7: What does working with AI actually change in someone's day? Ethan Mollick's framework: leadership, lab and crowd. This series follows the people, tests and choices behind adoption. Next: why people hide AI use. #AI #FutureOfWork
1
17
Part 3 of 3 - final: The goal is a new teammate who already knows how your team works. Skills make procedural knowledge transferable; they do not capture every kind of memory. #AI #AgentSkills #FutureOfWork
12
Part 2 of 3: A skill can be a folder your team versions and shares. The useful part: save the script once instead of asking the agent to rewrite it every time. Next: how shared skills help a team learn. #AI #AgentSkills #Workflows
11
Part 1 of 3: Would you choose a genius or an experienced professional to do your taxes? Anthropic’s case for reusable skills starts with that difference. Next: turn a folder into reusable expertise. #AI #AgentSkills #FutureOfWork
8
Reliability can mean removing decisions. Uber lets builders choose specific tools and fix parameters in advance, instead of asking the model to infer everything. #AgentReliability #MCP
9
A demo can be built overnight while a production system takes months. Dan Klein explains why reliability needs its own architecture, not just a stronger-looking prototype. STEP panel. #AI #BusinessOperations
7
One agent writes. Another critiques. Andrew Ng explains why the roles can differ even when the underlying model is the same. From his 2024 Sequoia talk. #AIWorkflows #CodeReview
9
Debug the instructions, tool descriptions and decision trail, not just the final answer. Barry Zhang suggests asking the model for its perspective, without replacing your own understanding. April 2025 talk. #AIAgents #Debugging
7
