Who Should You Trust to Teach You AI?

The episode opened with Perplexity Deep Research suddenly behaving very differently from the product Brian had used for months. Instead of detailed research, it returned short answers, mixed old conversations into new work and required far more effort to get a useful result. It was another reminder that AI workflows can break quickly when the underlying product changes.

Anne then shared how AI helped her small team keep two businesses operating while she stepped away from day-to-day work. The harder lesson was that useful automation required GitHub skills, clear SOPs, strict brand rules and basic data governance. A new nonprofit fundraising project made the stakes clearer because donor information and meeting recordings forced the team to decide where sensitive information could live before using AI.

The conversation shifted to AI model economics. Andy discussed pricing pressure on OpenAI and Anthropic from cheaper Chinese models, DeepSeek's reported use by hacking groups and concerns that anonymous models such as Ox Alpha can collect valuable user data during testing. NVIDIA's Groq technology added another angle, with new hardware reportedly producing thousands of tokens per second. The hosts also discussed whether businesses may accept slower local models when privacy matters more than speed.

The final section focused on the booming private AI education market, including a reported $19 million launch aimed at women in business. Anne argued that demand exists partly because corporate AI training often teaches tools rather than helping people rethink how work gets done. That led to a distinction between AI trainers and AI educators, with trust, change management and judgment becoming more important than simply showing people where to click.

Key Points Discussed

00:00:18 Episode Intro And The Road To Show 800

00:01:35 What Happened To Perplexity Deep Research?

00:07:40 Anne Returns And Shares Her AI Business Update

00:08:20 Moving A Small Business Toward Agentic Work

00:10:09 GitHub, Brand Rules And Model-Agnostic Operations

00:12:05 SOPs Let The Business Run Without The CEO

00:13:04 Data Governance Comes Before AI Deployment

00:18:03 Why Boring File Naming Still Matters

00:19:36 Andy Returns From Canada

00:21:23 OpenAI, Anthropic And The AI Pricing War

00:22:09 Are Chinese Models Driving Prices Down?

00:24:01 DeepSeek And AI-Enabled Cyberattacks

00:25:04 Is Ox Alpha Harvesting User Training Data?

00:26:57 NVIDIA Brings Groq Speed Into Its Hardware

00:28:26 AI Inference Reaches 3,400 Tokens Per Second

00:30:20 China, NVIDIA Chips And Export Controls

00:33:44 Privacy Versus Speed With Local AI

00:36:34 Private AI Education Becomes Big Business

00:37:01 The $19 Million AI Education Launch

00:38:02 Why Institutional AI Training Falls Short

00:39:58 Employees Become The AI Person Without Support

00:43:36 Trust Becomes The Moat For AI Educators

00:46:44 Are We Selling Spellcheck For A Typewriter?

00:49:36 AI Trainers Versus AI Educators

00:53:30 Setting Personal Rules For AI Use

00:54:23 AI Beauty Standards Become More Extreme

00:55:32 Episode Wrap-Up

The Daily AI Show Co Hosts: Brian Maucere, Andy Halliday, Anne Murphy, Beth Lyons