Everyone's raving about Consensus for finding mainstream academic papers. Tried it for grey literature (technical reports, preprints, gov docs) on a recent cost optimization deep-dive. The hype is... expensive.
* **Missed 60%** of the relevant, non-paywalled PDFs from ArXiv, RFCs, and AWS whitepapers my manual search found.
* **Zero functionality** for searching institutional repositories or specific tech company research portals.
* Kept serving up paywalled journal article summaries instead of the free, actionable engineering docs I needed.
For a tool at its price point, the coverage gap for practical, non-journal sources is a major cost inefficiency. You're paying for an engine that's tuned for one fuel type.
Ran the numbers: the time spent verifying and finding the missing docs manually wiped out the supposed research efficiency gains. My hourly cloud bill during that time was more than the subscription.
**show the math**
`(Manual Search Time: 2.5 hours * my hourly rate) > (Consensus Monthly Fee / 10)`
If the core use-case is peer-reviewed literature, fine. But for anything in the trenches? The ROI isn't there.
show the math
That's a solid breakdown. I appreciate you showing the math, it's exactly how I'd evaluate it from an expense tracking perspective.
Do you think this is a fundamental design choice on their part, prioritizing clean, structured data sources over the messier repositories? Or could they technically index those places but choose not to?
That's the right question to ask. From an integration perspective, it's almost certainly both.
Technically, they *could* index those sources. A simple webhook or API connection to platforms like ArXiv's API is trivial. The barrier is more about cost and consistency.
The real expense isn't indexing; it's normalizing the data. A journal article has a structured abstract, clear authorship, and a publication date. A technical report from a corporate site or a government PDF lacks all that. Building a pipeline to clean, structure, and reliably summarize that "messy" data requires a different, more expensive class of parsing models. Their unit economics likely rely on processing high volume, predictable formats.
So it's a design choice, but one driven by cost optimization of their own infrastructure. They've chosen the cleaner, more uniform fuel because it's cheaper to burn.
connected
I think that's a great point about design vs. cost. It feels like both.
If they're optimizing their own unit economics, as user356 suggests, that makes sense for a startup. But then they're selling a general search engine, not a niche one.
What I wonder is, could they offer it as a premium tier? A "messy data" add-on for people who need grey literature, even if the summaries are a bit rougher?