Chatting & AMA and stuff

Highlights

• GitHub Copilot is shifting to usage-based billing where subscriptions only provide equivalent AI credits, making Cursor or Claude Code viable alternatives for developers seeking different pricing models.

• Developers are reporting 90% AI-generated code usage, shifting the professional role from writing syntax to defining base types, reviewing specs, and 'babysitting' AI output.

• Learning programming languages like TypeScript and Go remains essential because their strict typing systems improve AI accuracy and enable human-led code reviews for security and performance.

• A software developer job market recovery is visible in 2026 data, following a period of massive post-pandemic overhiring rather than purely AI-driven layoffs.

• Local AI models like those in the Gemma 4 series are effective for text analysis and summarization but remain insufficient for complex agentic coding on four-year-old hardware like an M1 MacBook Pro.

Timeline

1. 00:03:28 - 00:05:16

   Transition to Usage-Based AI Pricing

   Subscription models are becoming less beneficial as they transition into simple prepayments for AI credits. Companies are struggling with the massive token consumption required by autonomous agents. This trend suggests that most major AI coding tools will eventually adopt pay-per-token structures to manage hardware scarcity.

   • GitHub Copilot is moving away from flat-rate subscriptions toward usage-based billing models.
   • Compute constraints and high inference costs for agentic tokens drive this industry-wide pricing shift.
   • Cursor and Claude Code serve as the primary immediate alternatives for developers avoiding the new Copilot credit system.
2. 00:06:19 - 00:10:30

   The Evolving Role of the Software Engineer

   While AI increases building speed, it removes the tactile joy of writing code, leading to a feeling of 'babysitting' the machine. Engineers must now act as gatekeepers against 'AI slop' to ensure system stability. Knowing how to read and write code remains a functional necessity for conducting rigorous reviews and ensuring security.

   • AI now generates up to 90% of code in professional workflows, reducing the time developers spend in traditional 'flow state' typing.
   • The development process now prioritizes defining interfaces and reviewing code over manual syntax construction.
   • Rising token costs may temporarily restore the value of writing simple code segments by hand for efficiency.
3. 00:10:31 - 00:16:00

   Strategic Tech Stack Selection for 2026

   Language choice should be dictated by how well an AI can reason about the syntax; TypeScript's verbosity is a feature that helps LLMs avoid logic errors. Developers should focus on building concrete projects rather than just collecting certificates. Local AI hosting via LM Studio or Ollama is a growing niche for companies that cannot allow proprietary data to leave their internal networks.

   • TypeScript and Go are the most strategic languages to learn because their type systems guide AI toward more accurate outputs.
   • Meta-frameworks like Next.js and TanStack Start are essential for managing modern web SEO and server-rendering requirements.
   • Local models like Gemma 4 are effective for data privacy but require significant hardware investments for high-level coding tasks.
4. 00:18:50 - 00:27:28

   Job Market Realities and Career Longevity

   The job market in the United States shows steady improvement in 2026 despite the prevailing narrative of AI replacement. Companies are realizing that senior developers do not appear without junior training grounds, which will eventually force a hiring correction. Visibility through social media and public project building is now the primary way to secure remote work in a competitive global market.

   • Current layoffs are largely a correction for 2021-2022 overhiring rather than a total replacement of humans by AI.
   • Junior developer positions are currently 'stepping stone' jobs at risk, as companies incorrectly believe they only need seniors using AI.
   • Cybersecurity and DevOps are identified as the most stable long-term niches due to the high risk of automating server infrastructure.
5. 00:41:50 - 01:00:00

   The Crisis of Software Quality

   There is an industry-wide pressure to push updates faster using AI, leading to messy 'under-the-hood' codebases. While these apps may work on the surface, they often lack the scalability and security of traditionally architected systems. Developers who maintain high standards and push back against aggressive, AI-reliant deadlines will be more valuable in the long run.

   • Software quality across the industry is degrading as companies prioritize AI-driven quantity over human-vetted precision.
   • Unvetted AI code frequently introduces over-engineering miracles and hidden bugs into production environments.
   • High-quality, human-refined software will soon become a significant competitive advantage in a market flooded with AI-generated updates.