Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Vivian Oliveres says Dwelverson processed about 1.75 million social posts a day and cost roughly €10 a day in inference, yet 800 cold emails produced just one trial. The mismatch is the point: finding possible buyers at scale is not the same as proving they want the product, or making it easy for them to act on a lead.
What Dwelverson was built to do
Oliveres describes Dwelverson as a solo-built B2B lead-intent product. It scanned Reddit, Bluesky, X, and Hacker News for people expressing an active need for software, then sent potential leads to vendors. The project reportedly began in February and shut down on September 15; the available author-published sources do not establish those dates’ year clearly enough to state it as verified.
The author’s post and post-mortem report the project’s operating figures and technical details. They are first-person accounts, not independently audited results or benchmarks that can be assumed to apply to other products.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchHow the pipeline handled the volume
Oliveres describes a 14-step cascade that moved from regex and embeddings through smaller encoders, a local LLM, and an API LLM. The approach used LLM-generated labels to train smaller encoders, which the author says cut inference costs by about 30-fold while retaining the same recall. The reported inference spend fell from an earlier estimate of about €100,000 per month to around €10 per day. These are project-specific figures, and the sources do not establish a common measurement period or independent cost audit.
The post-mortem also describes a frozen dataset, conditional recall, confidence intervals, nightly audits, and a golden cohort as parts of the evaluation process. Even with that measurement setup, the author says the task had an F1 ceiling of 0.27. That figure is the author’s account of this project’s task and evaluation, not a general model-performance result.
For compute, the system reportedly ran on an RTX 5090 kept on Oliveres’s desk. The post-mortem recounts months of GPU instability, including Xid 109 errors, driver segmentation faults, and swap livelock, before nightly runs stabilized. This is context about the project’s infrastructure, not a hardware test or recommendation.
Rank #2
Why processing posts did not translate into customers
Some sources mattered far more than their volume suggested
According to Oliveres, Reddit made up only 0.2% of collected posts but about 80% of true leads. That concentration meant the pipeline’s total volume was a poor proxy for the value of its results: the small slice with the strongest reported lead yield came from one platform.
Recommended Free Tools
The author also says clients could not readily reply to leads from accounts without enough platform history, citing karma and Contributor Quality Score. That describes the author’s experience; it should not be read as a comprehensive or current statement of Reddit policy. A lead can be relevant and still be difficult for a buyer to contact in the place where it appeared.
Outbound outreach did not establish buyer demand
Oliveres reports that 800 cold emails generated one trial. A separate outreach effort aimed at Reddit self-promoters sent 60 emails and received no replies. The author attributes the first result to sender reputation and spam filtering, but presents that as an interpretation rather than a measured diagnosis.
Those outreach figures do not show whether the underlying leads were accurate, whether the offer was compelling, or whether a different sales channel would have worked. They do show that the project’s reported ability to find and classify posts did not, by itself, create a reliable path to trial users.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The lesson the author drew
Oliveres’s retrospective advice was to test the market before building the full system: find one potential lead manually and see whether a buyer wants the result. In the author’s words, “What I’d do differently fits in one line: test the market in February, with one lead found by hand, before writing the pipeline.”
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →That sequence separates two questions the project’s technical work could not answer on its own: can a system identify useful intent, and will a specific buyer pay for or act on those leads? A manual example tests the second question before the cost and complexity of automation are sunk.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
