Cohere Compass 2 vs Perplexity Comet

“The primitive here is a retrieval pipeline that can ingest mixed-format documents — PDFs with embedded charts, scanned tables, image-heavy slides — and return semantically relevant chunks without requiring a preprocessing ETL step per modality. That's a real problem: anyone who's tried to build RAG over a 10,000-document enterprise library knows the pain is 80% in the ingestion layer. The DX bet is that Cohere handles the multimodal parsing so you don't glue together a PDF parser, a table extractor, and a vision model yourself. The on-prem deployment option is actually the headline feature for the buyer, not the multimodal part — that's what gets it past legal review. My skip concern is documentation: the blog post is long on capability claims and short on API surface, schema design, and what 'image understanding' means at query time versus index time. Show me the query API, then we'll talk.”

“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 direct competitors are Azure AI Search with multimodal indexing, AWS Kendra, and increasingly any RAG stack bolted onto GPT-4o's native PDF vision. Compass 2's real differentiator is not the multimodal capability — every major cloud provider is shipping that — it's the on-premises deployment for enterprises with data residency requirements, combined with a retrieval model trained specifically for enterprise document retrieval rather than general web content. The scenario where this breaks is at the 'chart understanding' claim: interpreting a bar chart semantically in a way that survives a specific quantitative query ('find all documents where Q3 revenue exceeded Q2') is a much harder problem than the blog post implies, and I've seen this class of tool hallucinate chart data confidently. What kills this in 12 months isn't a competitor — it's that the chart and table comprehension doesn't hold up under production query loads and the feature gets quietly deprioritized. I'm shipping it narrowly: for text-heavy PDFs with some visual elements in air-gapped environments, this is probably the best available option right now.”

“Comet is competing directly with Arc's Browse, Google's Project Jarvis, and Anthropic's computer-use demos — except those shipped broadly and Comet is invite-only for a Q3 2026 general rollout. The specific failure scenario is obvious: any task requiring login state management, CAPTCHAs, or multi-domain auth handoffs falls apart immediately, and Perplexity hasn't shown evidence of solving those problems at scale. My prediction for what kills this in 12 months: Google ships Gemini-native browser automation in Chrome, erasing Comet's differentiation with zero distribution disadvantage. To earn a ship, Comet needs to demo booking a multi-leg international flight with seat selection, payment, and confirmation — live, unscripted, first try.”

“The buyer is a VP of IT or Chief Data Officer at a regulated enterprise — financial services, pharma, government — and the budget comes from the data infrastructure or compliance line, not a software tools budget. That's a real check-writer with a real problem: they have document libraries they legally cannot send to OpenAI's API, and they need search that works across formats. The on-prem deployment option is the actual moat here, not the multimodal capability — Cohere has been building that distribution channel for two years and it creates genuine switching costs once it's integrated into an enterprise's document management stack. The risk is that the pricing model is 'contact sales' all the way down, which means a long sales cycle and high CAC that has to be recovered on large contracts. What survives the model-gets-cheaper scenario is the enterprise integration layer and compliance certifications, not the retrieval model itself — Cohere needs to be pricing for that, not for compute.”

“The buyer here is the $20/mo Perplexity Pro subscriber, which means Comet is a retention feature masquerading as a product launch — there's no incremental revenue attached to it unless Perplexity spins it into a higher tier. The moat question is brutal: Comet's agentic capability sits on top of browser automation infrastructure that Google, Microsoft, and OpenAI are all building simultaneously, and none of them need to charge $20/mo to distribute it. The specific business problem is that Perplexity is spending engineering capital on a browser at exactly the moment when its search revenue model remains unproven — this is a distraction bet that only makes sense if it dramatically increases Pro retention or unlocks enterprise contracts. What would need to change: a dedicated Comet tier at $40-50/mo with verifiable task-completion SLAs and an enterprise sales motion.”

“The thesis Compass 2 is betting on: enterprise knowledge is fundamentally multimodal — it lives in slide decks, scanned contracts, financial tables, and annotated diagrams — and the first retrieval system that treats those formats as first-class citizens rather than edge cases will own the enterprise search layer. That's a plausible and falsifiable bet, but the dependency is that 'understanding' a chart means something semantically useful at query time, not just 'we embedded the image.' The second-order effect that matters here isn't faster document search — it's that if this works, structured data that currently lives locked in PDFs becomes queryable without a data engineering team to extract it, which shifts power from BI teams who own structured pipelines toward anyone with a document library. Cohere is riding the trend of on-premises LLM deployment for regulated industries — that trend is real and accelerating, and they're on-time to it, not early. The future state where this is infrastructure is 'every regulated enterprise has a Compass instance the same way they have an Active Directory instance.' I'd believe that in five years if the chart comprehension claim is real.”

“The thesis here is falsifiable and specific: by 2028, the browser is not a viewport but an execution environment, and the team that controls the AI-browser layer controls the intent graph of the web. Comet is betting on this at the infrastructure level — not bolting agents onto a tab, but rebuilding the browser around the agent primitive. The second-order effect that matters most is what this does to web analytics and SEO: if agents complete tasks without humans seeing pages, the entire attention economy built on pageviews collapses. Comet is riding the computer-use trend line and is roughly on time — OpenAI Operator launched earlier, but browser-native execution versus API-layer automation is a real architectural distinction worth watching. The dependency that has to hold: agentic task completion rates must cross ~85% reliability before mainstream users tolerate it.”

“The job-to-be-done is sharp: complete a web task I would otherwise do manually across 4-8 browser tabs. That's a real, recurring job with measurable time cost, and Comet is one of the first products to attempt it at the browser layer rather than the script or extension layer. The onboarding concern is real though — invite-only access means the vast majority of Pro subscribers can't evaluate whether this replaces their current workflow, making it impossible to call this a complete product today. The opinion baked into Comet is correct: the browser should understand goals, not just URLs. The gap between what's shipped and what's needed is a public availability date that isn't six months away, and documented task success rates so users can set realistic expectations before switching.”