Reviewing Hot Takes From Legendary Web Devs

Theo - t3․gg

Tweet Performance and Community Response

Initial tweet asking for "spiciest takes about web dev" reached 4+ million views
Generated almost 1,000 hot take replies
Analysis covers biggest/most engaging responses (not necessarily best)

DKP (X-State creator) take: UI components wrong place for app logic most of the time
- Easy/convenient initially but creates long-term maintenance issues
- Makes understanding app flow difficult
Counter-perspective: Component-level logic has clarity/compartmentalization benefits
- Don't need to understand entire app to work on small pieces
- Global state increases implementation/maintenance overhead for small features
- Context: Effectiveness vs. whole-app understanding trade-off
- Recommendation: Use global state responsibly, avoid 15+ context providers

Take: Tailwind wouldn't be as popular if React had easy component-scoped styling
Analysis: React needed styling solution, drove multiple approaches
- CSS modules, styled-components, Stitches, Material UI attempts
- Tailwind won as best componentization solution
Tailwind benefits beyond React:
- Makes CSS more enjoyable/efficient to write
- Eliminates system-wide style relationship concerns
- Provides consistent cross-project experience
- Functions as design system through utility classes
Comparison to inline styles: Tailwind = optimized inline styles with design system benefits

Devon (Adobe React Aria) observation: Huge mobile/tablet traffic but poor touch consideration
Hover behavior problem: No hover equivalent on mobile devices
- Tooltips, darker states on hover don't work
- Touch users get degraded experience
Call to action: Web developers must improve touch user experience
Links provided for accessible button design and hover interaction best practices

DKP follow-up take: Enormous capital/talent/effort spent on marginal web app speed improvements
Diminishing returns: Industry far into performance optimization curve
Bundle size obsession: Arguing over 80kb vs 20kb bundles when apps will write 100kb+ anyway
Missing focus: Should optimize big application performance, not hello-world kilobyte counts
Device reality: Devices not improving as fast, but code not getting slower proportionally
Recommendation: End performance wars, appreciate current state

Parasocial take: "Next.js is not that great"
Historical value: Revolutionary for pushing React devs to consider servers
- Challenged single-page-app-only mentality
- Forced front-end/back-end relationship thinking
App Router benefits: Best developer experience for back-end/front-end integration
- Enabled rapid development of complex applications
- Marker Thing and Upload Thing built efficiently
Ecosystem context: Not only solution, but unique combination of:
- Quality experience + wide adoption + company support
- Better bet than theoretically superior but less adopted alternatives
Industry influence cycle: Innovation flows from cutting-edge to mainstream frameworks

Jack Harrington: "Never roll your own auth" (cautious approach)
Kent C. Dodds: "Most web apps should not use auth service"
Current reality assessment: Auth services increasingly necessary
- Next Auth doesn't support App Router well yet
- Clerk currently best option for modern tools
- Lucia Auth shows promise in open source space
Service advantages:
- Eliminates edge case debugging time
- Enables multi-platform support (mobile + web OAuth sharing)
- Handles JWT, Edge runtime, mobile complexity
Cost/benefit: $50/month vs. full-time team of 3+ engineers
Bottom line: "Auth is hard; if you think auth is easy, your auth is bad"

Dax controversial take: JS ecosystem janky due to lack of talent, not tech
- Claims mediocre engineers historically, good ones finally arriving
- Previous leadership "couldn't get package manager right"
NPM historical context: Pre-NPM had no better package managers
- Revolutionary at launch time
- Nuget, Pacman, Nix not developer-focused solutions
- Enabled open source application development
Alternative framing: JS lowered engineering bar to ship working products
- Enabled larger developer population
- More applications serving user needs
- Average JS dev quality lower due to scale, not selection
Rust vs. JS developer comparison: Rust developers self-selected enthusiasts vs. JS developers often job-assigned
Ecosystem strength: JS enables more things to be built, despite quality variations

Rocks take: "99% of software is trivial, building not hard, getting people to care/pay is hard"
Scaling evolution: Previously difficult scaling now much easier
Future prediction: Shift from few 100k-person companies to many 5-10 person companies
- Building perfect solutions for smaller, specific user bases
- Sufficient revenue for small team salaries
Code value perspective: Code is cheap, should be thrown away
- User experience is valuable
- Tools should enable better user understanding
- Building is not the hard part

Common advice criticism: "Learn HTML/CSS/JS basics before frameworks"
Problem with "learn basics first": No clear completion criteria
- Could spend year studying CSS and still not know everything
- New developers don't know when they've learned "enough"
Alternative approach: Start by building specific things
- Clear success criteria (working application)
- Users understand applications better than language specifics
- Learn fundamentals when needed, not preemptively
Framework benefits for beginners: Svelte shows HTML/CSS/JS in context
- React has more industry usage and resources
- Both enable quick application building

Controversial take: "90% of web dev industry could disappear, quality of life would improve"
Strong disagreement: Industry growth enables specialized solutions
Small team capability: 5-person teams can build solutions for specific problems
- More specialized, quality solutions possible
- Better than 15 companies serving everyone
Personal example: Marker Thing + Lossless Cut workflow
- Web app extracts Twitch timestamps to CSV
- JavaScript-based video editor processes results
- Enables efficient video production workflow
Niche solutions impact: Average person has specialized interests
- All niche interests better served by current ecosystem
- Specialized tools solve specific problems effectively
Industry evolution: Moving toward more companies serving fewer users with better solutions

Try for free