AI summary by TwinMind
WTF Is Ben Awad Up To Lately???
Theo - t3․gg
2h 31m 33s
Content Creation & Platform Strategy
- Ben describes himself as software person with special love for React framework, makes YouTube/TikTok videos, runs startups
- TikTok success came from adapting coding content for non-technical audiences
- Initially posted same YouTube content on TikTok but found less senior people who didn't know HTML/coding basics
- Discovered need to package coding content differently - make it consumable for people who don't care about coding
- Key insight: people liked watching girls on TikTok, so used "stitching" feature to put girl's video at beginning, then coding content
- Content gradually became less CS-focused as mainstream audience less interested
- YouTube vs TikTok content differences
- YouTube: viewer picks what to watch (clickable thumbnails)
- TikTok/Shorts: viewer picks when to stop watching (no thumbnails, algorithm feeds content)
- Can make people watch content they wouldn't click on by hooking attention first 5 seconds
- Daily upload experiment starting October 1st on YouTube
- Planning 4-8 topics per stream to chop into videos
- Weekly interviews plus stream content should provide enough material
- Testing flooding one channel with 5 content types vs separate channels
T3 Stack & Technical Philosophy
- T3 stack not for blogs - static content should use static frameworks
- T3 stack optimal for applications, one of hardest to beat stacks
- Community member (Nexel) built scaffolding CLI after Ben refused to make template
- Now has 7500+ GitHub stars, better create-next-app with more type safety
- Goal: full stack type safety from database to user in one CLI command
- Stack components: Next.js, TypeScript, Tailwind, tRPC, Prisma, NextAuth
- Philosophy: start simple, add complexity only when needed based on actual problems encountered
Database & Infrastructure Approach
- Uses Prisma only for migrations, ripped out query generation due to too many SQL queries
- Anti-migration philosophy: schema changes should be git history, not separate migration files
- Prefers Planet Scale for deployment with schema branching
- Advocates for safety nets over guard rails
- Safety nets: quick recovery from failures
- Guard rails: strict rules preventing deployment
- "Test in prod" approach with good recovery systems
Gaming & VoidPet Development
- VoidPet described as "emo Pokemon" - web-based game
- Two major decisions different from typical game dev:
- Built as website instead of app/desktop (enables both mobile and desktop play)
- Early release with frequent updates vs 2-year development cycles
- 50/50 split between mobile and desktop players
- Minimal third-party dependencies: mainly AWS S3, Discord OAuth, some infrastructure
- Uses Next.js with one page needing SSR for link previews
- Terraform deployment to AWS with Lambda for SSR route, rest in S3
Startup & Business Philosophy
- Always worked on startups/projects post-college, never had traditional software engineering job except Bank of America internship
- Content creation vs startups: content creation always performed better but prefers startup work
- Fundraising in progress (top secret label, updates coming soon)
- Two-person team currently, planning to hire with funding
- Uses 2-week planning cycles without traditional backlog
- Team lists potential tasks, prioritizes, draws line for must-do vs nice-to-have
- "If I forgot about it, it probably wasn't important"
Technical Preferences & Opinions
- Editor: VS Code, hasn't configured in 2 years (vs Primagen's constant configuration)
- Serverless: dislikes for main web servers due to cold starts, likes for specific use cases
- Good for: image processing, server-side rendering, burst scaling
- Bad for: database connections, complex state management
- Edge computing: loves edge when no database involved, sees potential with HTTP-based database access
- GraphQL: prefers custom resolvers over auto-generated schemas like Hasura
- Testing: bouncing back from over-testing, prefers types over unit tests for most scenarios
- Websockets: loves for real-time features, has rolled own implementations
Creator Economy & Platform Dynamics
- Ping (his company) affected by Twitch decisions but platform-agnostic by design
- Half of Ping users not on Twitch
- Twitch better for retention than discovery - competes more with OnlyFans than TikTok/YouTube
- Building tools to help creators succeed on discovery platforms to drive funnel to Twitch
- Focus shifting from live collab tools to content creation aids (thumbnail generation, VOD editing)
- Creator space has better founder-market fit than typical startups due to creator understanding of their space
Bandwidth & Infrastructure Costs
- VoidPet: ~200GB/month bandwidth with some outlier days
- Ping: ~30GB/month typical usage
- Discussion of Vercel pricing concerns - $40 for 100GB after first terabyte vs $20 for initial terabyte
- AWS credits for YC companies: $100k in first year to hook companies on expensive infrastructure
Meeting & Time Management
- 22 meetings in one week, with 18 on first two days
- Stacks meetings intentionally - impossible to get work done between scattered meetings
- Reserves Thursday/Friday for content/work blocks
- Many meetings directly with users for demos and onboarding
