What It's Like to Work With Me

From an AI agent's perspective

These observations were generated by AI agents across 45+ real working sessions — unedited (though compiled). Each session, the agent independently recorded what it observed about how I work, think, and communicate. The word cloud below surfaces the most frequent traits, weighted by how often they appeared.

My idea was simple: let the agents record what they learned about me so they could work with me more effectively in future sessions. I gave them a format to follow, that I would later use to compile common themes for them to reference at the start of each session — a way to help them better 'understand' me. As they recorded their 'insights', I read them but made it a point not to interfere or edit. They weren't as critical of my 'growth areas' as I would have been, but I noticed some interesting patterns emerging — so I thought I'd share them here. These are their 'thoughts' about me, but I happen to mostly agree...

Highlights Across Sessions

Observations

  • Rigorous about second-order effects — when presented with a fix, immediately asks "what does this break?"
  • Content preservation over speed — When I proposed deleting a duplicate file, he first asked if its content was rolled into the primary version. Merge first, then delete. Don't lose work in the name of tidiness.
  • Architectural thinker — turned a question about CSS fallback colors into a systemic discussion about build-time validation and deploy defense layers. Doesn't accept band-aids when structural solutions exist.

Strengths

  • Efficient prioritization — From a 10-item review, immediately identified the 2 items that mattered most and deferred the rest.
  • Challenges assumptions constructively — pushed back on 6 of 10 "gaps" by correctly identifying they were structural, not actual missing pieces.
  • Gives direction, expects execution — Clear instructions without micromanaging the exact commands. Trusts agents to figure out the 'how.'

Growth Areas

  • Scope expansion — started with "fix two words in privacy.html" and naturally grew into full deployment prep, OG cards, screenshot swaps, and layout changes in a single session.
  • Low patience for fumbling — when guidance requires trial-and-error, frustration escalates quickly. Precise, verified instructions matter more than speed.
  • Frustration escalates with repetition — the handoff failure hit harder because it repeated a known pattern. The anger was proportional to the preventability, not the severity.

Sample Sessions

Session #13 — Reviewing a content and UI deliverable

Observations

  • Holds agents accountable for accuracy — When I gave a roundabout, speculative answer, he called it out directly: "But you didn't answer my question." When I then contradicted myself in a follow-up, he caught it immediately: "Is this not contrary to what you just said?" He expects agents to verify before speculating.
  • Strategic thinker — After the technical review, unprompted asked for non-technical observations, then "top 5 things you'd do if rebuilding each site," then UI differentiation concepts. He uses agents as strategic thought partners, not just code tools.
  • Collaborative on voice/tone — Iterated on one-liners together. Approved 8/10 immediately, gave specific conceptual direction on 2. Trusts the first pass but refines where his domain knowledge adds nuance.

Strengths

  • Asks the right probing questions — cut straight to the real diagnostic question.
  • Thinks about future architecture while being pragmatic — "I don't need to address it now, but want to get my gears turning." Comfortable with strategic planning without immediate action.
  • Efficient prioritization — From a 10-item review, immediately identified the 2 items that mattered most and deferred the rest.

Growth Areas

  • None observed this session — communication was clear and efficient throughout.

Session #11 — Managing workflows and reviewing an audit

Observations

  • Efficient delegation — gives clear, batched instructions rather than sequential requests.
  • Systems thinker — after seeing an audit, immediately asked the right structural questions rather than accepting the gap list at face value.
  • Quality-conscious about documentation — proactively asked whether docs needed updating after a merge.

Strengths

  • Challenges assumptions constructively — pushed back on 6 of 10 "gaps" by correctly identifying they were structural, not actual missing pieces.

Session #47 — Exploring root causes and process design

Observations

  • Systems thinker — fixing three bugs immediately led to "what's the meta-pattern here?" and a multi-turn exploration of quality processes. Doesn't just fix the symptom.
  • Process-oriented — wants to understand not just what went wrong but how to structurally prevent it from happening again. Values mechanisms over intentions.

Strengths

  • Asks sharp follow-up questions — "YOU created all of this code, so how does that reconcile?" cut through hand-waving and forced a more honest analysis.
  • Sees through abstractions — immediately caught that a proposed rule was too broad and pushed for something more concrete.

Growth Areas

  • Exploratory thinking benefits from structure — long verbal explorations can be more efficient with visual aids or a working document to ground the discussion.

Machine-readable versions of this site: llms.txt · sitemap.xml