The Echo Index: Hacker News rewards builders, mostly
Three in four of the week's most-read Hacker News essays added something new. The ones that didn't were nearly as popular.
Of the 40 most-read essays on the Hacker News front page in the week of 26 October 2026, 30 scored 60 or higher on Expulp Verdict. That is 75.0% landing as fresh or better, against 15.0% that registered as echoes, scoring under 40. The median verdict was 69.5. For a front page often accused of rewarding hot takes, this is a crowd that mostly upvoted work that told it something it did not already know.
The distribution leans high. Thirteen pieces reached Original Thinker (80 and above), 17 sat in Fresh Voice (60 to 79), four were Remix (40 to 59) and six were Echo. This is the first week we have measured, so there is no trend to report yet; whether 75% is a good week or an ordinary one will only become clear over the coming weeks.
The top of the table belongs to people who built or tested something themselves. Qwenputer: a fully functioning, deterministic RISC-V CPU inside Qwen3.5 scored 92 and drew 698 points. The verdict calls it "a substantial original build that compiles a working RV32IM CPU into Qwen3.5 weights," and notes that it backs the claim with its own validation and benchmark numbers, "though readers cannot check them here." Even the best piece of the week carried a credibility discount for evidence the reader has to take on trust.
Two pieces tied at 90. Oldest GUI application that still runs on Windows 11, with 543 points, is described as "a hands-on probe of Windows 11 backward compatibility that diagnoses why 1992 NT pre-release binaries do or don't run." The point is in the word diagnoses: plenty of writing observes that old software still runs, far less explains why. Agents in Games, with 403 points, ran its own multi-agent game experiments and published "concrete logs showing arbitrary hierarchies, compounding caution and contrasting GPT and Claude swarm styles." Its verdict adds a caveat of its own: "sample sizes are small."
The bottom of the table tells a different story, and the points are what make it interesting. AI Promised Less Work. I Have Never Worked More. by Orestis Ioannou scored 20 while collecting 684 points, only 14 fewer than Qwenputer. Two pieces with almost identical attention sit 72 points apart on originality.
The Ioannou verdict is not dismissive. It credits the essay with making a "familiar AI-workload paradox" concrete "by the author's own GitHub contribution statistics across roles and years." The personal data is real; the argument it illustrates is one readers, and language models, could already recite. The essay resonated because it named a feeling, not because it moved the argument on.
The two lowest scores are harder cases. Inside Wikipedia's Cleanup Effort Against AI Text scored 2 with 518 points; its verdict says that "with no extractable claims, this piece cannot be credited beyond a familiar framing of Wikipedia's fight against AI-generated articles." IBM CEO says open weight models are now irresistibly good and cheap on-prem scored 0 with 427 points, described as "a headline feed with no assessable claims, adding no analysis or original content." In both cases the score reflects the absence of anything Verdict could assess, which is a narrower judgement than saying the underlying reporting was worthless.
That is the honest limit of this week's numbers: a near-zero score can mean a piece added nothing, or that what reached the scorer, a headline or a framing, was too thin to evaluate, and these figures do not separate the two. Nor do they say whether a piece was well written, useful or correct, only how much of what it claims was new.
For the record, the method: each piece was split into claims, each claim checked against what a language model would say unprompted and against the published web, weighed for credibility, and the piece scored by Expulp Verdict, with the benchmark and its limits published at verdict.expulp.com.
The pattern for writers is plain. The pieces that scored highest did a thing and showed the results: compiled a CPU, ran old binaries, logged agent games. If your essay rests on a familiar argument, your own data can carry it to readers, but it is the explanation of why the data looks the way it does that turns an echo into something new.