Approximately one year ago, it was easier for venture capitalists to fund AI companies that were still searching for an idea due to the novelty of AI and the abundance of untapped verticals.

Bob McGrew, one of the originators of the forward-deployed engineer model at Palantir, is reportedly skeptical of its current overuse and believes it should be applied sparingly.

The early Bitcoin market participants were reportedly enraged by Coinbase's approach of implementing KYC procedures and collaborating with traditional financial institutions.

Upon its launch, OpenAI received mostly negative press and faced extreme skepticism from the established AI research community, who doubted that a young team could achieve AGI.

DoorDash entered the food delivery market when it was already crowded with established players like Postmates, Grubhub, and Seamless.

DoorDash's contrarian strategy was to build a pure software marketplace for food delivery, rejecting the popular "full stack" playbook of the time (circa 2014) where competitors like SpoonRocket and S...

The environment for founding AI startups has become more competitive, as many verticals like insurance and banking now have multiple startups, and there has not been a major, market-shaking AI model r...

The most successful companies to emerge from the mobile era, such as Uber, DoorDash, and Instacart, were based on non-obvious ideas that were not widely predicted when the iPhone launched.

OpenAI's practice of crawling the entire web for training data is considered to be in a legal gray area, with arguments for it being either fair use or massive copyright theft.

Elon Musk's idea to build reusable rockets for SpaceX was considered "blasphemous" and impossible by many rocket scientists at the time.

GigaML competes against Sierra and Decagon in the customer support AI space by offering a product that works better out-of-the-box, resulting in a faster sales and onboarding process.

Google's Gemini Pro 2.5 and Gemini Flash 2.5 models are considered to be equivalent in capability to OpenAI's competing models.

Exa.ai established a scale economy moat by making a large, early capital investment to crawl a significant portion of the web before AI agents became mainstream.

Despite Google's established brand and user base, OpenAI's ChatGPT has more daily consumer users than Google's Gemini.

Legora is counter-positioning against early legal AI leader Harvey by focusing on building a better application-layer product rather than on model fine-tuning.

The speaker Harj believes Google Gemini's grounding API, which uses the Google index for real-time information, is more accurate than Perplexity.

The speaker Harj switched to using Google's Gemini as his primary model in 2025, finding it superior for reasoning tasks.