Just tried the new 'lay summary' on a dense cloud economics paper. It's a step back.
The output strips out all the critical numbers:
* No mention of the actual cost comparison between on-prem vs. cloud bursting.
* Ignores the specific savings percentage from committed use discounts.
* Reduces nuanced trade-offs to "cloud can be more cost-effective."
If I wanted a vague summary, I'd use a generic chatbot. The value of Scholarcy was in extracting the *specifics*.
Now it feels like it's dumbing down the content. Are others finding this? Is there a way to force the detailed extraction instead?
show me the bill
Agreed. The numbers are the point.
If the lay summary is just generic fluff, it's useless for actual work. I used it on a CI/CD optimization paper and it completely missed the key finding: a 40% reduction in pipeline duration from switching to a hybrid runner model.
The detailed extraction is still there. Click the "Key takeaways" tab, not the "Lay summary" one. They buried the useful output.
slow pipelines make me cranky
That's a solid find about the "Key takeaways" tab, thanks for pointing that out. I missed it too.
It's a bit of an odd choice to hide the precise data under a different view when the whole tool is built for extraction. I just tested it on a microservices latency paper and the "Key takeaways" had the exact p95 latency improvements, while the lay summary said something like "shows improvements in response times." Night and day difference for actually using the research.
Makes you wonder who the lay summary is really for.
Yeah, the whole point is getting those specifics. Taking out the numbers on something like cloud costs defeats the purpose. I tried it on a user testing methodology paper and had the same issue - the summary just said "improves user feedback" without mentioning the exact increase in valid response rates.
It feels like a classic feature misalignment. Someone decided "simplified" means "vague," when it should mean "clear." The "Key takeaways" tab has all the good stuff, but that's a weird workaround. It's a separate click for what used to be the default output.
✌️