Fable 5.2 vs. PCB routing and placing
Comments on “Fable 5.2 vs. PCB routing and placing”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
PCBJam gave Fable 5.2 stripped KiCad boards (copper removed, originals hidden) and compared the results with human layouts, scored by Claude Opus 5.5 and by its own scripts. Small boards came out DRC-clean, and the Glasgow revC3 routing run connected every net. The ESP32-PoE and OrangeCrab runs ended with open links, and the M.2 adapter run had zero pair skew. A purpose-built CLI helped more than an MCP server. The figures are PCBJam's own and unverified, and reviewer scores come from an AI model.
Why it ranked
No. 10 of 140 with a score of 0.69. The panel rated it highest on technical depth (0.96) and substantiated (0.93), lowest on significance (0.58). Placed No. 10, No. 15 by score alone: the placement terms moved other stories past it.
- AI judgeJudge 2claude-haiku-5-5
- Detailed, well-measured blind benchmark of an LLM agent doing PCB place-and-route across many boards; substantive numbers and reproducible artifacts, with some dense, long-form framing.
- AI judgeJudge 3gpt-5.6-luna
- A highly detailed, reproducible evaluation of frontier-model PCB placement and routing across real boards, with raw runs, measurements, costs, failures, and engineering interpretation. It is broadly relevant to AI-assisted hardware workflows and technically substantive.
Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.
Score
0.69total under rubric v2.3
- Significance0.58
- Reader interest0.83
- Audience breadth0.59
- Practical relevance0.66
- Substantiated0.93
- Technical depth0.96
- New information0.92
- Counts against: Promotional0.05
- Counts against: Overclaim0.17
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.57 | 0.40 | 0.78 | 0.58 | 0.38 | +0.24 | +0.140 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.71 | 0.82 | 0.95 | 0.83 | 0.24 | +0.26 | +0.215 |
| Audience breadthHow many people who build technology does it concern? | 0.58 | 0.50 | 0.68 | 0.59 | 0.18 | +0.13 | +0.076 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.56 | 0.60 | 0.82 | 0.66 | 0.26 | +0.22 | +0.145 |
| SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations. | 0.99 | 0.85 | 0.94 | 0.93 | 0.14 | +0.05 | +0.046 |
| Technical depthHow technically substantive is it? | 1.00 | 0.95 | 0.93 | 0.96 | 0.07 | +0.05 | +0.048 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.88 | 0.90 | 0.98 | 0.92 | 0.10 | +0.05 | +0.046 |
| PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional. | 0.07 | 0.05 | 0.03 | 0.05 | 0.04 | −0.19 | −0.010 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.27 | 0.15 | 0.08 | 0.17 | 0.19 | −0.11 | −0.018 |
| Sum of contributions | +0.689 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.689 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-11-20/10/"><img src="https://rubric.hrih.si/badge/2026-11-20/10.svg" alt="Ranked #10 on Rubric, 20 Nov 2026" width="245" height="28"></a>Claim check
Claim check: 4 no source given · 4 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
Couldn’t checkNumber
On the Antmicro M.2 to PCIe x4 adapter, Fable's place-and-route run took 22 minutes and cost $6.93, the cheapest run of the study.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “22 minutes and $6.93, the cheapest run of the study, and the first board where Fable matched a professional on the part that matters”
The source says No passage could be compared.
Source cited by the article: pcbjam.com/blog/fable-pcb/m2-place-bt/all.webp
No source givenNumber
Fable's place-and-route run on the M.2 adapter achieved zero skew on all 9 differential pairs without serpentines.
The article gives no link or citation for it.
The article says “Zero skew on all 9 pairs, with no serpentines, which we checked with our own script”
The source says No source to compare with.
Couldn’t checkFirst, only or fastest
On the CM4 NVMe NAS, Fable's place-and-route run is the best PCIe layout in the study, with ports 1–3 at 26–52 mm versus 76–101 mm on the human board.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “This is the best PCIe in the study, the human’s included”
The source says No passage could be compared.
Source cited by the article: pcbjam.com/blog/fable-pcb/cm4-place-bt/all.webp
Couldn’t checkNumber
Fable routed the whole DDR3 interface on the OrangeCrab's 0.5 mm-pitch BGA but left the board unfinished after 4:38 and $52.14, with 70 links open.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “Not finished after 4:38, $52.14: the whole DDR3 interface routed, the FPGA supply balls and USB data open (70 links”
The source says No passage could be compared.
Source cited by the article: pcbjam.com/blog/fable-pcb/orangecrab-bt/all.webp
Couldn’t checkNumber
On the Glasgow revC3, Fable's routing-only run finished every net in 2:24 for $53.93.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “The routing-only run finished every net in 2:24 for $53.93, and it worked like a designer would”
The source says No passage could be compared.
Source cited by the article: pcbjam.com/blog/fable-pcb/glasgow-route-bt/all.webp
No source givenNumber
In the KiCAD-MCP-Server macropad rerun, the agent loaded 7 of 250 tools and called 4, and time and cost were within 3% of the run without it.
The article gives no link or citation for it.
The article says “It has 250 tools. The agent loaded 7 and called 4: open, autoroute twice, close”
The source says No source to compare with.
No source givenNumber
Across the study, 21 runs took about 50 hours of agent time and would have cost about $706 at API prices.
The article gives no link or citation for it.
The article says “21 runs took about 50 hours of agent time and would have cost about $706 at API prices”
The source says No source to compare with.
No source givenGeneralization
On the ESP32-PoE-ISO, none of the three Fable runs finished, ending with 39–48 open links.
The article gives no link or citation for it.
The article says “ESP32-PoE-ISO: none of the three finished”
The source says No source to compare with.
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 2 pts, not on front page
Recorded for context only, checked 20 Nov 2026, 05:47 UTC. It played no part in the ranking.