Raw Data with Rob Collie

The episode has a genuine core insight about subscription-to-per-token pricing as a deliberate strategic trap, and the development-vs-deployment cost split is useful for operators. However, the density is low - much of the runtime is scheduling chat, pilings anecdotes, and meandering speculation rather than actionable observations.

“The whole time they were sitting on just an absolute ace card, which is we'll just declare the newest version of our things to be not covered by subscription.”

“I do think this also kind of uniquely precious in the development space...you're developing software, you're developing websites, you're developing new systems and that's where a lot of intensive token usage happens.”

The 'checkmate' framing of Anthropic's subscription strategy is a moderately fresh take, and the OpenAI 'fake high ground' competitive angle shows some independent thinking. But the discussion stays largely reactive to news rather than building a first-principles argument, and no contrarian position is really defended.

“The whole time they were sitting on just an absolute ace card, which is we'll just declare the newest version of our things to be not covered by subscription.”

“Or they can choose to temporarily fake like they're taking the high ground and say, oh, we think that's dirty pool...when they're really saying, we think bait and switch comes a little bit later.”

Both participants are genuine practitioners who daily-drive these tools and are building real products, which gives their usage data credibility. However, there is no actual guest - it is two co-hosts from the same small firm - and neither brings executive-scale deployment experience or research depth that would elevate the conversation beyond informed hobbyist territory.

“I've had a skill in claw that I've been using for, gosh, probably over a year now called daily planning.”

“The Pathfinder product that I've mentioned to you that we're building. We had this big milestone feature set that we want to get through”

The personal usage cost estimate ($200/month subscription vs. $6-7k/month at API rates) and the website audit example with 20 findings and one false positive are the most concrete data points and genuinely useful. The rest of the episode trades in speculation and rough analogies rather than named benchmarks, pricing tiers, or hard business metrics.

“Oh, probably six to seven...But we're small, right? We're just like a two. Like we're a small company. I would imagine for P3. Like, I could see so times that number by 50”

“There's maybe 20 key findings and maybe one of them was bad.”

Rob does ask for a concrete example ('All right, give me an example') and circles back to a missed follow-up from a prior episode, which shows some intentionality. But the conversation is almost entirely two people agreeing with each other, no claim is challenged, and questions rarely push past the surface of a topic before the thread drifts.

“All right, give me an example.”

“I regret it afterwards, not asking you about that. But, like, in what ways did you perceive Opus 4.7 to be wacky?”