AI tools and performance: New benchmarks, vibe-coding, and business applications emerge
What happened
Three stories emerged around AI tools and business performance: a startup lawyer at Synthesia used the company's avatar software to build an AI version of himself, a researcher named Peter Gostev launched "BullshitBench" to test whether AI models can detect nonsensical questions, and Fast Company offered practical AI project ideas for solo business owners.
How it was covered
Both outlets — Business Insider and Fast Company — framed AI as a practical, accessible tool for professionals. Business Insider led with novelty: a lawyer who "vibe-coded" his own AI avatar, plus a researcher testing whether models can spot BS (Google Gemini 3.0 "struggles," per the excerpt). Fast Company took a how-to angle, noting that "most people use AI like a search engine" and pushing solopreneurs toward deeper use cases. The tone across both outlets is enthusiastic and utility-focused, with no critical framing of AI risks or hype.
What one side told you that the other didn't
Business Insider's BullshitBench piece adds a rare skeptical data point — a benchmark specifically designed to expose model weaknesses, with a named model (Google Gemini 3.0) failing. Fast Company's excerpt doesn't name any model failures or limitations at all, keeping the framing squarely aspirational.
Why They Framed It This Way
Business Insider balanced novelty with a credibility hook — the BullshitBench story gives tech-savvy readers a critical lens while the vibe-coding story delivers the entertaining human angle. Fast Company's audience skews toward small business owners hungry for actionable advice, so the "here's what to try" frame serves readers looking for a productivity edge, not a debate about AI reliability.
What To Watch Next
BullshitBench is the story with the most legs here — if Gostev publishes full model rankings or Arena adopts it as a standard benchmark, it could pressure Google and others to respond publicly. Watch whether Gemini's team acknowledges the result or whether other benchmarking bodies pick up the methodology. Track Gostev's publication timeline and any official responses from Google DeepMind in the next 48 hours.
Get this analysis every day
Signal/noise aggregates 100+ sources across the political spectrum so you can see how different outlets cover the same story — free.
Sign up free — it's daily