developer experience

Bilingual

Examples are the best documentation
2 months ago
- Most developers need a single example to understand documentation, but official sources rarely provide them.
- Formal docs target experts, yet many developers juggle multiple contexts and need quick context restoration.
- Python's max() doc example shows complexity: users must understand *, /, positional-only, iterables, keyword-only, key, etc., while a simple code example would suffice.
- Community projects like clojuredocs.org show the value of practical examples that include related functions, making them indispensable for daily coding.
Things I Don’t Know as of 2018 — overreacted
2 months ago
- Experienced engineers often face unrealistic expectations to know everything, but knowledge gaps are normal.
- The author lists many technologies they don't know, including Unix, low-level languages, containers, microservices, Python, and more.
- Admitting knowledge gaps does not devalue experience; even seasoned developers have valuable expertise despite gaps.
- The author encourages sharing knowledge gaps to combat impostor syndrome and unrealistic expectations in the tech industry.
So, I tried to build an Android app
2 months ago
- Setting up a Google Play developer account involves submitting government ID, a utility bill, and a $25 fee, but new accounts risk automated suspension even after payment.
- New Google accounts can be suspended unexpectedly, forcing developers to use personal accounts with more history to avoid instant rejection.
- Google's closed beta testing requires joining a Google Group, clicking a separate link to become a tester, and waiting 1–48 hours for the app to appear, unlike Apple's simpler TestFlight system.
- Google Groups used for testing can be suspended without content or posts, further complicating the testing process.
- Since November 2023, new Play Store accounts must have at least 12 closed beta testers, wait 14 days, pass a quiz, and maintain testers throughout; older accounts are grandfathered in.
- The 12-tester requirement penalizes indie developers and encourages fake testers or buying accounts, rather than improving app quality.
- Complex installation steps cause users to abandon testing after initial attempts, leading to low engagement and frustration.
- The author's app was available on the App Store in 2–3 weeks but still not public on Google Play, discouraging future Android development.
IDEs are done, and they are not coming back
2 months ago
- The author created a list of 30 developer experience issues in JavaScript tooling upon joining VoidZero in January.
- A friend's question led the author to realize they no longer needed both Sublime Text and their IDE, motivating them to remove Sublime Text.
- The author stopped writing code by hand months later, making the previously identified paper cuts irrelevant and forgotten.
- The author has now deleted their todo list and plans to delete VS Code, declaring that IDEs are obsolete and will not return.
The Human-in-the-Loop Is Tired
2 months ago
- LLMs make programming genuinely useful but also destabilizing, leading to burnout if ignored.
- Development with LLMs introduces 'fatigue of supervision': reviewing AI-generated code lacks the satisfaction of traditional programming and feels lonely.
- AI increases work intensity and parallel task initiation but doesn't improve thoughtful completion, creating a 'human reward function problem'.
- LLM-assisted programming shifts skills rather than eliminates them, emphasizing taste, nuance, and architectural expertise as key differentiators.
- New skills are emerging, like pre-mortem analysis and encoding expertise into LLM instructions, showing evolution rather than death of craft.
Software Has Changed
2 months ago
- Software development has changed drastically and quickly, leaving experienced engineers shell-shocked.
- Development is now faster but not necessarily better, focusing more on quantity over quality.
- Flow states have been replaced by 'thinking...' states, and code is often discarded without thought due to cheap experiments.
- Good developers can now do more, using saved time to review, polish, refactor, and add extra features.
- Technical debt can be addressed before it reaches the main codebase, improving overall code quality.
- The field has changed overnight, requiring adaptation to build faster and think at a higher level to avoid tunnel vision.
- A key question arises: if the end result is solid, thoroughly reviewed, and trusted code, does the creation process truly matter?
The gauge broke: devs felt 20% faster with AI, measured 19% slower
3 months ago
- A controlled trial found experienced developers felt 20% faster but were actually 19% slower when using AI tools in familiar codebases.
- AI speeds up typing, which isn't the bottleneck for experts, and adds overhead like prompting, waiting, and reviewing often subtly wrong outputs.
- Team-level data shows increased pull request volume and size but no net delivery gain, with more code churn and less refactoring, indicating verification has become the new bottleneck.
- Tool builders are shifting focus to agent-first IDEs, emphasizing verification over generation, aligning with the identified bottleneck in development workflows.
- The feeling of speed is misleading; successful teams must measure actual production outcomes and restaff review stages, rather than relying on perceived velocity.

developer experience

Examples are the best documentation

Things I Don’t Know as of 2018 — overreacted

So, I tried to build an Android app

IDEs are done, and they are not coming back

The Human-in-the-Loop Is Tired

Software Has Changed

The gauge broke: devs felt 20% faster with AI, measured 19% slower