Source-derived protocol · AI Engineer
Agent Output Review
When you have agent output to review, use a review kit that shows the entire changes and the changed files so you can go file by file, get visual evidence such as videos or photos of what changed, and run a few agent instances in parallel on the same task to pick the best implementation.
4 actions
AI Engineer's principle
A review kit that surfaces the entire changes, the changed files, and visual evidence lets you quickly understand agent output.
AI Engineer's example
- Ido Salomon describes creating a review kit that gives him the entire changes, shows which files changed so he can go file by file, and can surface visual evidence — videos or photos of what changed — so he can quickly understand what he wants.
This is what the creator describes doing. It is not assigned to you.
Your version
4 total
Review agent output by inspecting the full changes file by file, using visual evidence, and comparing parallel implementations.
From this source · Adapted for your goal
The review steps are the creator's; the schedule for applying them is yours to set.
Use a review kit that shows the entire changes and the changed files so you can go file by file.
Recurring · on your cadenceDone when: You have gone through the changed files in the review kit.
Get visual evidence — videos or photos of what changed — to quickly understand the changes.
Recurring · on your cadenceDone when: You have visual evidence of what changed.
Run a few agent instances in parallel on the same task and pick the best implementation.
Recurring · on your cadenceDone when: You have picked the best implementation from the parallel agent instances.
Review whether this review process still fits the agent work you are doing.
Review · on your cadenceDone when: You have decided whether the review process still fits your current agent work.NuthinButta scaffolding
Why it works
What AI Engineer says this builds. Their framing, not ours — every line links to where they said it.
Raise the ceiling but also lower the floor: bring ~90% of people into the agent future without burnout
Ido Salomon argues we need to raise the ceiling but also lower the floor — bringing around 90% of the people in the world into this future without having them burn out.
View in the full analysis →Agents' growing capability calls for a lower-denominator, accessible interface
Ido Salomon argues that Warcraft-style complex UIs can be daunting for non-technical users, so a lower-denominator interface is needed now that agents are more capable — trying to find what makes stuff accessible.
View in the full analysis →Humans are the bottleneck in agent-driven work
Ido Salomon argues that in reality the human is the bottleneck in agent-driven work: each agent must be steered, directed, and reviewed, and supervising many agents at scale is exhausting and can lead to burnout. He frames the needed skills as ones people already have from outside programming.
View in the full analysis →
Source-derived protocol
Based on “We're the bottleneck, but we don't have to be — Ido Salomon, MCP Apps” by AI Engineer.
- Source fidelity verifiedPassed · Nov 10, 2026
- Protocolability reviewedPassed · configurable template
- Editorially reviewedNot yet
- Supported by independent sourcesNot evaluated
Source fidelity means each creator-worded element was checked against the exact statements it cites. It says nothing about whether the approach works.