This week on Fresh From the Labs, Shilpa, Kevin, and Jared dive into a fresh wave of AI model and product drops. First, they unpack the arrival of Kimi K2, a massive trillion-parameter open-weight model from Moonshot AI that shows compelling performance on coding benchmarks and could be run locally.

Next, they discuss Grok 4, which has claimed top spots on key benchmarks but comes with an awkward launch and a steep price tag, sparking a conversation about pricing, user lock-in, and whether benchmarks translate to real-world utility.

The team also explores Perplexity's new AI-powered browser, Comet, and what it signals for the future of web browsing and competition in a space that major players like OpenAI are also rumored to be entering.

The conversation then pivots to a crucial and surprising topic: the misalignment of expectations around AI efficiency. Referencing a recent study, the hosts debate the finding that developers feel more productive with AI tools but may actually be slower. This leads to a nuanced discussion on the steep learning curve, the skill required to effectively use AI, and the pressure on companies and individuals to adopt these tools, even if it doesn't immediately boost productivity. Join us for a candid look at the latest models and the hard realities of integrating AI into daily workflows.