notthecode
@notthecodeblog.bsky.social
What makes great software? (Hint: it's not just code.) Exploring the human elements behind exceptional tech. #NotTheCode
https://notthecode.com/
https://notthecode.com/
Read the full post:
notthecode.com/the-art-of-l...
notthecode.com/the-art-of-l...
How to Lead Without Doing: A Guide for Tech Leads
You didn’t stop contributing when you stepped into leadership—you just started building something less visible and far more impactful: the environment where great work happens.
notthecode.com
December 23, 2025 at 3:35 AM
Read the full post:
notthecode.com/the-art-of-l...
notthecode.com/the-art-of-l...
The goal is to design the machine so it runs without you. That’s not absence. That’s architecture.
December 23, 2025 at 3:35 AM
The goal is to design the machine so it runs without you. That’s not absence. That’s architecture.
A rule that hurts (but helps): if you’re still the best coder on the team, you’re designing a bottleneck.
December 23, 2025 at 3:35 AM
A rule that hurts (but helps): if you’re still the best coder on the team, you’re designing a bottleneck.
Hiring + retention is leverage. Career growth is maintenance. Psychological safety is the merge gate for bad news.
December 23, 2025 at 3:35 AM
Hiring + retention is leverage. Career growth is maintenance. Psychological safety is the merge gate for bad news.
But the team doesn’t need you to be the emergency coder. They need you to fix delivery flow: clarity, unblockers, decisions, dependencies.
December 23, 2025 at 3:35 AM
But the team doesn’t need you to be the emergency coder. They need you to fix delivery flow: clarity, unblockers, decisions, dependencies.
The Hero Trap is real: you jump back into PRs because it’s the only work that still gives you a clean “done.”
December 23, 2025 at 3:35 AM
The Hero Trap is real: you jump back into PRs because it’s the only work that still gives you a clean “done.”
If you’re a new EM, your editor being empty can feel like failure. It’s not. Your output just moved out of Git.
December 23, 2025 at 3:35 AM
If you’re a new EM, your editor being empty can feel like failure. It’s not. Your output just moved out of Git.
Engineering managers don’t stop building. They stop building code and start building the system that produces it.
December 23, 2025 at 3:35 AM
Engineering managers don’t stop building. They stop building code and start building the system that produces it.
Full post: notthecode.com/psychologica...
If your team is feeling the “clean diff / shallow review” squeeze, send this to your reviewers and pick one rule to adopt this week: intent + what wasn’t verified + failure-path walkthrough.
If your team is feeling the “clean diff / shallow review” squeeze, send this to your reviewers and pick one rule to adopt this week: intent + what wasn’t verified + failure-path walkthrough.
Psychological Safety in Teams for AI Auto-Complete
AI autocomplete increases output, not understanding. This post reframes psychological safety in teams as technical honesty under velocity—intent-first reviews, explicit verification, and practical PR ...
notthecode.com
December 18, 2025 at 5:04 AM
Full post: notthecode.com/psychologica...
If your team is feeling the “clean diff / shallow review” squeeze, send this to your reviewers and pick one rule to adopt this week: intent + what wasn’t verified + failure-path walkthrough.
If your team is feeling the “clean diff / shallow review” squeeze, send this to your reviewers and pick one rule to adopt this week: intent + what wasn’t verified + failure-path walkthrough.
Psychological safety in teams is permission to be technically honest under deadline pressure: ask the awkward questions while the change is still cheap.
December 18, 2025 at 5:04 AM
Psychological safety in teams is permission to be technically honest under deadline pressure: ask the awkward questions while the change is still cheap.
If the author can’t trace the failure path, don’t merge it. That’s not being harsh. That’s how you avoid shipping a professionally-formatted mystery.
December 18, 2025 at 5:04 AM
If the author can’t trace the failure path, don’t merge it. That’s not being harsh. That’s how you avoid shipping a professionally-formatted mystery.
High standards changed. Syntax quality is cheap now. The expensive part is intent + constraints + verification. If a PR can’t answer “what did you not verify?”, it’s not done.
December 18, 2025 at 5:04 AM
High standards changed. Syntax quality is cheap now. The expensive part is intent + constraints + verification. If a PR can’t answer “what did you not verify?”, it’s not done.
New failure mode in AI-assisted teams: output velocity outruns understanding velocity. The diff gets bigger, the review gets shallower, and “LGTM” becomes a coping mechanism.
December 18, 2025 at 5:04 AM
New failure mode in AI-assisted teams: output velocity outruns understanding velocity. The diff gets bigger, the review gets shallower, and “LGTM” becomes a coping mechanism.
AI makes code look “reviewed” even when nobody understands it. Clean diff + green tests is not the same thing as a shared mental model. Psychological safety is what lets someone say: “pause, I need to trace this.”
December 18, 2025 at 5:04 AM
AI makes code look “reviewed” even when nobody understands it. Clean diff + green tests is not the same thing as a shared mental model. Psychological safety is what lets someone say: “pause, I need to trace this.”
Full playbook + prompts (including review + TDD tips): notthecode.com/your-first-a...
AI Pair Programming: Not Your Code Vending Machine
Stop treating AI like a code vending machine. This playbook helps you use tools like GitHub Copilot and ChatGPT. These tools act as personal mentors. They help you learn faster, work with confidence, ...
notthecode.com
December 12, 2025 at 8:22 AM
Full playbook + prompts (including review + TDD tips): notthecode.com/your-first-a...
Use a simple loop: ask for thinking first (risks/tradeoffs), then request minimal code with reasoning, then do the “explain-it-back” test. If you can’t explain it to a rubber duck, delete it.
December 12, 2025 at 8:22 AM
Use a simple loop: ask for thinking first (risks/tradeoffs), then request minimal code with reasoning, then do the “explain-it-back” test. If you can’t explain it to a rubber duck, delete it.
Better model for AI pair programming: you are always the Driver. AI is the Navigator. It can suggest, critique, list risks—never decide.
December 12, 2025 at 8:22 AM
Better model for AI pair programming: you are always the Driver. AI is the Navigator. It can suggest, critique, list risks—never decide.
The bill shows up later: you can’t explain what you merged, you go quiet in reviews, and when prod breaks at 2 AM you’re debugging “someone else’s” code.
December 12, 2025 at 8:22 AM
The bill shows up later: you can’t explain what you merged, you go quiet in reviews, and when prod breaks at 2 AM you’re debugging “someone else’s” code.
Stop using AI like a code vending machine: paste error → accept code → ship. It feels fast. It’s a trap.
December 12, 2025 at 8:22 AM
Stop using AI like a code vending machine: paste error → accept code → ship. It feels fast. It’s a trap.