Realistic advice about software dev right now

TL;DR

Breaking into software dev is harder than ever because AI inflated apparent competence while job scarcity collapsed urgency, so likability and genuine passion now matter more than credentials.

Key Points

• 1.The hiring equation has fundamentally changed. Jobs now require demonstrating all three factors: urgency (company-controlled), likability (shared), and competence (mostly yours) — whereas in 2016, high urgency alone could carry below-average candidates like Theo into $125k roles.
• 2.Supply/demand math disadvantages new grads. With roughly 100k CS grads competing for ~70k new tech jobs, a ~30k surplus exists — meaning the bottom 25% who previously got hired by accident largely cannot anymore.
• 3.Interview processes are inaccurate by nature. Amazon's 1–4 rating system (designed to force above/below-median decisions) still produced panel disagreements of two full levels, meaning candidate placement has a ~1–2 standard deviation error range.
• 4.Theo got hired at Twitch despite bombing the interview. He froze on language choice, failed C++ problems, failed front-end questions, and only got a 3-month contract because the team liked him and had no better options in 2016.
• 5.AI has broken traditional competence signals. GitHub profiles, portfolio quality, and number of shipped projects are now unreliable hiring signals because AI lets incompetent devs appear skilled, hurting genuinely capable devs who can't yet prove it.
• 6.Dunning-Krueger traps new devs in false confidence. Early learners feel expert-level confident; as knowledge deepens, self-assessed skill plummets — the opposite pattern (chronically underestimating real skill) is imposter syndrome, which plagues many strong devs in Theo's community.
• 7.Being the smartest dev in your circle is a career liability. Theo argues that lacking access to more-capable developers — common for isolated or remote devs — stunts growth because you lose the measuring stick needed to calibrate your actual skill level.
• 8.Virtual community can substitute for in-person networks. Since most viewers can't afford SF's ~$4k/month median rent, embedding in online communities (Discord, Twitter/X, live streams) replicates the calibration and mentorship benefits of a strong local tech scene.
• 9.The 'who' question is Theo's greatest career hack. Instead of reading code, he prioritizes finding the creator — their history, motivations, and background — and making genuine low-ask contact, which is how he built relationships with figures like Ryan Carneato (SolidJS creator).
• 10.Use AI to go deeper, not to get answers. Ask AI 'why' and 'how should I think about this' rather than 'do this for me'; letting it give one small unblocking hint preserves the learning process that builds real, demonstrable competence.