AI & Tech: Special-ed AI plans, coding skill loss, and Gemini Flash leading benchmarks

A daily brief on what the world is saying. PodSnacks tracks the top 200 podcasts and distills the most important conversations into a fast, high-signal listen. AI adoption in special education, concerns over lost debugging skills among developers, and the latest on Google Gemini models and video tools. A focused briefing on how these shifts are reshaping classrooms, coding practices, and model performance. Sources (episodes in this brief): Could artificial intelligence improve special education?: https://podsnacks.com/media?id=07e2f9ad267a4adba675ba630e31a2d7 https://podsnacks.com/media?id=1ac78afe712b4eb0bba487049da4c9fb TNB Tech Minute: SpaceX Filing Launches Countdown to IPO: https://podsnacks.com/media?id=2432ee9321f24211b3baf43230a72655 OpenAI Is Making Plans to File to Go Public Very Soon: https://podsnacks.com/media?id=5fc6cb72ca6b4ce0ae43b27ba0b643d3 Nvidia Earnings Lead to Share Slump; SpaceX IPO Details: https://podsnacks.com/media?id=628f88410879474295a3a69865d4f92a Has AI Conquered Coding? (It’s Not So Simple…) | AI Reality Check: https://podsnacks.com/media?id=70ab62cb3bc94fafb4fc2caf855efae6 Why Google Isn't Chasing Claude Code: https://podsnacks.com/media?id=c38774923cac48edac549516a5c4c2dc

Transcript

SPEAKER_00 0:05

Podsnacks! The news you need from the voices you trust. This is PodSnacks. Here's what's driving the conversation. Thursday, May 21st, 2026. Over half of special education teachers used AI to draft individualized education plans last school year. One California middle school chatbot turned test data into usable IEP sections in roughly 10 minutes instead of hours. Responsible districts now require mandatory training and human review to avoid privacy or accuracy violations when AI writes full plans. Consider this from NPR highlighted how these tools are speeding up paperwork while raising new oversight questions. Developers report losing core debugging skills after relying on agentic coding tools. Many now call for writing 20 to 100% of important code manually and supplying pseudocode rather than letting agents handle unfamiliar work. Reddit threads titled Losing My Ability to Code Due to AI Show Juniors Hitting the Industry Junior Year Wall when they skip foundational struggle. Deep questions with Cal Newport explored why that deliberate practice still matters. Google's new Gemini Flash 3.5 model leads OS world benchmarks, yet consumes 3.5 times more tokens than GPT-5 5 Medium. The company's OmniVideo model adds character-consistent editing and scene changes for shorts creators. Overlapping tools such as Spark, Anti-Gravity, and AI Studio risk confusing users despite 900 million monthly Gemini users. The AI Daily Brief noted the performance gains alongside the added complexity for everyday creators. Nvidia reported $81.6 billion in quarterly revenue, up 85% year over year, driven by data center demand. The company announced an $80 billion buyback and raised its dividend 2,400% while guiding $91 billion for the next quarter. It is shifting reporting into data center and edge computing segments and shipping zero hopper products to China this quarter. Bloomberg Daybreak covered how these moves reflect the continued scale of AI infrastructure spending. Cerebras went public at a $64 billion valuation after securing a multi-billion dollar open AI inference contract and avoiding HBM and 3 nanometer bottlenecks. The move shows how specialized hardware players are carving out space even as larger chip ecosystems expand.