site: deck downloads on the homepage, and retract the takum accuracy lead - #857
Merged
Conversation
…lead Adds two CTAs to the investment block linking the deck as a PDF in English and in Russian, served statically from /deck/. Retracts 2.1x/2.6x. The figure was produced against an oracle that negated wrongly on every negative code (fixed in trinity-fpga #606) and at a nominal rather than an equal-stored-width budget. Re-measured at equal stored width the result is a three-way tie (research/TEKUM_VS_TNF_LINE_2026-08-19.md): 16-bit tekum10 8.561e-3 takum16 5.697e-3 TNF(4,8) 5.323e-3 32-bit tekum20 1.434e-7 takum32 1.264e-7 TNF(4,24) 1.203e-7 The number is kept visible in the invest facts under the existing `retracted` tag rather than deleted, so a reader who saw it can see it retired. The same correction was sent to Hub71 in writing on 2026-08-20, and leaving it published here while telling them it was withdrawn was the contradiction this closes. The 66 LUT fact drops "at 974.66 MHz": that figure is an unconstrained nextpnr critical-path estimate (the logs read PASS at 12.00 MHz), so it does not belong on a card tagged `measured` next to an area count that does. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
github-actions Bot
added a commit
that referenced
this pull request
Aug 21, 2026
site: deck downloads on the homepage, and retract the takum accuracy lead (#857) Adds two CTAs to the investment block linking the deck as a PDF in English and in Russian, served statically from /deck/. Retracts 2.1x/2.6x. The figure was produced against an oracle that negated wrongly on every negative code (fixed in trinity-fpga #606) and at a nominal rather than an equal-stored-width budget. Re-measured at equal stored width the result is a three-way tie (research/TEKUM_VS_TNF_LINE_2026-08-19.md): 16-bit tekum10 8.561e-3 takum16 5.697e-3 TNF(4,8) 5.323e-3 32-bit tekum20 1.434e-7 takum32 1.264e-7 TNF(4,24) 1.203e-7 The number is kept visible in the invest facts under the existing `retracted` tag rather than deleted, so a reader who saw it can see it retired. The same correction was sent to Hub71 in writing on 2026-08-20, and leaving it published here while telling them it was withdrawn was the contradiction this closes. The 66 LUT fact drops "at 974.66 MHz": that figure is an unconstrained nextpnr critical-path estimate (the logs read PASS at 12.00 MHz), so it does not belong on a card tagged `measured` next to an area count that does. Co-authored-by: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Two changes to
apps/website/src/content/tnf.ts, the homepage content module.1. Deck downloads. Two CTAs in the investment block, linking the deck as a PDF in English and in Russian. The files are already live and served statically from the site repo:
Both build from source (
business/hub71/build_deck.py,build_deck_ru.py). The Russian one uses this site's own design system: palette and φ-scaled type and spacing lifted fromassets/index-*.css, origin-label wording and colours from the site bundle, Manrope substituted for Outfit because Outfit carries no Cyrillic.2. Retract 2.1×/2.6×. The accuracy lead over takum is withdrawn, in the FAQ, in
limits, and in the invest facts.Why the retraction
The figure was produced against an oracle that negated wrongly on every negative code (fixed in trinity-fpga #606), and at a nominal rather than an equal-stored-width budget. Re-measured at equal stored width,
research/TEKUM_VS_TNF_LINE_2026-08-19.md:No format wins by more than the width it gives up.
tnf_paper.tex:6473states it directly: "is not an accuracy advantage at equal storage — there is none at 16 and 32 bits."The number is kept visible in the invest facts under the existing
retractedtag rather than deleted, so a reader who saw it can see it retired.The same correction was sent to Hub71 in writing on 2026-08-20 as part of an application correction. Leaving it published here while telling them it was withdrawn was the contradiction this closes.
Also
The
66 LUTfact drops "at 974.66 MHz". Everyws_gfternary_*.logreads974.66 MHz (PASS at 12.00 MHz)— the constraint inbench.xdcnever bound, so that is an unconstrained nextpnr critical-path estimate, not timing closure. It does not belong on a card taggedmeasuredbeside an area count that is one.Not addressed here
974.66/827.81/0.1797still appear elsewhere in the module (hero, frontier). Those are the site's headline numbers and how to frame them is the author's call, not a mechanical fix — but thelimitssection currently does not disclose that the frequency column is unconstrained, and it should.🤖 Generated with Claude Code