Perplexity Comet vs Sup AI

“The direct competitors here are Arc with Browse, Dia, and honestly just Operator from OpenAI — which already does agentic browser automation and has the distribution advantage of the most-used AI brand in the world. Comet's specific failure scenario: any workflow that requires logging into accounts with 2FA, handling CAPTCHAs, or navigating SPAs with dynamic state — which is most of the interesting automation targets. My 12-month prediction is that OpenAI or Google ships 80% of this natively into their existing browsers and Perplexity's differentiation collapses to 'we also have a search box.' To earn a ship, Comet needs to demonstrate agent reliability rates on real-world tasks above 80%, not cherry-picked demos.”

“Extraordinary claims require extraordinary evidence. A 7.41 point jump on HLE via ensembling — without publishing methodology — smells like benchmark gaming. The latency of running 339 models in parallel is also a real concern for anything other than async research tasks.”

“The thesis here is falsifiable: by 2028, the browser becomes the agent runtime rather than a document viewer, and the team that owns the browser layer owns the automation stack. The dependency is that OS-level agent APIs from Apple and Microsoft don't make the browser layer irrelevant before Comet builds distribution. The second-order effect nobody's talking about is that if this works, Perplexity gains clickstream data on user intent that no search engine currently has — not just queries but the full task graph, which is a training data moat. They're riding the trend of intent-layer consolidation and they're early enough that the category isn't defined yet, which is the right time to plant a flag.”

“Model ensembling is an underexplored direction in the race to reduce hallucination. If Sup AI's approach scales, it could be more durable than fine-tuning individual models — you get the wisdom of the crowd across model families, training data, and architectures simultaneously.”

“The primitive is: a Chromium fork with an injected agent that can read and manipulate the DOM plus call Perplexity's inference API. The DX bet is that bundling the runtime into the browser eliminates the permission and injection problems that plague extension-based agents — that's actually the right call architecturally. But the moment of truth is trying to automate something that matters to you specifically, and without a published automation scripting interface, a local action log, or any developer surface to inspect what the agent is actually doing, this is a black box. The weekend alternative for a competent engineer is Playwright with a function-calling loop, which gives you full observability. Until Comet ships an agent trace viewer or a scripting API, it's a consumer demo, not infrastructure.”

“The HLE claim needs independent verification, but the underlying ensemble approach is architecturally sound for factual Q&A tasks. Running 339 models is expensive — pricing will be the gating factor for production use. The $10 free credit is a fair trial.”

“The buyer here is the Perplexity Pro subscriber who already trusts the brand with search — this is a land-and-expand move and the expand story is actually credible because browser replacement has natural stickiness once your bookmarks and session history are in. The pricing is smart: Comet ships included with Pro, which lowers the adoption friction to zero and lets Perplexity study task completion data before charging for the feature separately. The moat question is real though — the switching cost of a browser is high but Perplexity doesn't own an OS, a mobile platform, or an enterprise SSO, so enterprise expansion is a hard road. The business survives model commoditization because the value is in the task graph and user behavior data, not the inference itself.”

“For creative work, ensemble outputs tend to regress toward the mean — you get the most-agreed-upon version of something, which is usually the least interesting version. This is a tool for factual accuracy, not creativity. I'd stick with a single strong model for writing.”