From e3cb6efb8217debef32737074efcdd1898c87c0b Mon Sep 17 00:00:00 2001 From: Andrius Mikonis <1989882+foxis@users.noreply.github.com> Date: Mon, 17 Aug 2026 00:06:03 +0000 Subject: [PATCH] content: strip process artifacts from published articles MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Published articles must read as first-person work. Several carried leftover production scaffolding, two of them visible to readers on the live site. Removed, both languages where applicable: 1. meteor75-pro-ii-fighting-resonances/index.md — DRAFT NOTES FOR ANDRIUS block, which opened with "delete this comment block before publishing" and was published anyway. Six numbered items including "still yours to confirm" and "needs your review". 2. meteor75-pro-ii-fighting-resonances/index.lt.md — VERTIMO PASTABA and ANDRIUI PERŽIŪRĖTI block, including "nesprendžiau nė vieno, palikta tau" ("I resolved none of them, left to you"). Checked before deleting rather than after: the LT note claimed the dual-stated numbers were left unresolved, but the LT body does carry the resolving section ("Metodo pastabos, kurias verta pasilikti", "Kaip skaityti skaičius"), and all eleven disputed figures appear consistently in both language bodies. The note was stale, not load-bearing. The stray "wrong for the fifth time" it mentioned existed only inside the comment itself, and the snake-chart-data.json it referenced lives outside this repo, so no live reference broke. 3. colorimetry/laser-power-meter/index.md — two "TODO: Add board.jpg" and "TODO: Add software-ui.jpg" comments. 4. fpv/drone-detector-nn, both languages — reader-visible figure placeholders. "*[TODO: Training curve ...]*" and "*[TODO: 10x10 confusion matrix]*" were italic markdown, not comments, so they rendered on the published page inside the Results section. Also removed the paired IMAGE comments. 5. fpv/betaflight-vtx-indexing-trap, both languages — "[TODO: confirm the exact TH3 V02 unlock sequence before publish.]" rendered live, publicly admitting the instruction above it was unverified. Not simply deleted, because deleting it would leave an unconfirmed claim reading as confirmed. Converted into an explicit statement in the author's voice: the exact unlock sequence was not confirmed on his own unit, so the button-hold is the general pattern rather than a recipe. Deliberately left alone after checking each in context, because they are legitimate authorial content and not scaffolding: - dml-first-steps: "solvable with small mid/high-range speakers on my TODO list" is ordinary prose. - esp32-as-bluetooth-audio: the "# TODO" heading is a genuine future-work section written in first person. - edgetx-cockpit-voice-audio-and-sharing: "Before publishing a config, check for and scrub" is the subject of the post. - about/me.lt.md and rate-profiles-persistent-stats.lt.md: "peržiūrėti" is normal Lithuanian prose. Verified on a local Hugo 0.163.3 build: no artifacts remain in content prose, all four affected pages render clean in both languages, clean in an unminified build too, and article bodies are otherwise unchanged. Two follow-ups for the author, deliberately surfaced rather than hidden: drone-detector-nn is still missing its training-curve chart and confusion matrix, and laser-power-meter is still missing board.jpg and software-ui.jpg. --- .../colorimetry/laser-power-meter/index.md | 3 -- .../betaflight-vtx-indexing-trap/index.lt.md | 2 +- .../fpv/betaflight-vtx-indexing-trap/index.md | 2 +- .../content/fpv/drone-detector-nn/index.lt.md | 4 -- blog/content/fpv/drone-detector-nn/index.md | 6 --- .../index.lt.md | 22 ----------- .../index.md | 39 ------------------- 7 files changed, 2 insertions(+), 76 deletions(-) diff --git a/blog/content/colorimetry/laser-power-meter/index.md b/blog/content/colorimetry/laser-power-meter/index.md index 7b516bd..14d2389 100644 --- a/blog/content/colorimetry/laser-power-meter/index.md +++ b/blog/content/colorimetry/laser-power-meter/index.md @@ -137,8 +137,6 @@ $$ P_{cal} = \frac{V^2}{R} $$ - - ### The Circuit I use a [Wheatstone bridge](https://en.wikipedia.org/wiki/Wheatstone_bridge) configuration with the two thermistors as variable resistors. The bridge converts the resistance difference (temperature difference) into a voltage difference. @@ -259,7 +257,6 @@ Where $s$ is the measured slope. The quadratic term accounts for nonlinearities I connected the RP2040 to my computer via USB serial. A Go application displays the differential signal in real-time and lets me trigger calibration measurements. - ![Software interface showing the measurement signal](software-ui.jpg) The software: diff --git a/blog/content/fpv/betaflight-vtx-indexing-trap/index.lt.md b/blog/content/fpv/betaflight-vtx-indexing-trap/index.lt.md index c67f957..ed36a8f 100644 --- a/blog/content/fpv/betaflight-vtx-indexing-trap/index.lt.md +++ b/blog/content/fpv/betaflight-vtx-indexing-trap/index.lt.md @@ -187,7 +187,7 @@ Perkrauk, ir ratukas daro tai, ką siūlo fizinis pasukimas: apačia = žema, vi Sutvarkyk indeksavimą, ir 400 mW *turėtų* pasirodyti ratuko viršuje. Jei vis tiek nepasirodo, gali būti, kad juos laiko pats VTX. Daugelis SmartAudio VTX — tarp jų ir TH3 — parduodami su **apribotu (regionui užrakintu) režimu**, ribojančiu išvestį dėl reglamentų. Tokioje būsenoje aukštos galios įrašai tavo `vtxtable` yra apibrėžti teisingai, išrenkami teisingai — ir tiesiog atmetami aparatinės įrangos. -Apribotas režimas gyvena VTX'e, ne Betaflight'e, tad jokie CLI pakeitimai jo nepašalins. Jį išjungi ant paties įrenginio — paprastai palaikant VTX mygtuką (arba atitinkama SmartAudio „unlock" komanda), kad išeitum iš regionui užrakinto profilio. [TODO: prieš publikaciją patikrinti tikslią TH3 V02 atrakinimo seką.] +Apribotas režimas gyvena VTX'e, ne Betaflight'e, tad jokie CLI pakeitimai jo nepašalins. Jį išjungi ant paties įrenginio — paprastai palaikant VTX mygtuką (arba atitinkama SmartAudio „unlock" komanda), kad išeitum iš regionui užrakinto profilio. Tikslios atblokavimo sekos savo TH3 V02 nepatikrinau, tad mygtuko palaikymą vertink kaip bendrą principą, o ne kaip receptą, ir konkrečią seką tikrink savo VTX dokumentacijoje. Požymis diagnostinis: jei žema ir vidutinė galia išrenkamos teisingai, bet vien viršutinis lygis lieka miręs po indekso pataisymo, įtark apribojimą, ne konfigūraciją. diff --git a/blog/content/fpv/betaflight-vtx-indexing-trap/index.md b/blog/content/fpv/betaflight-vtx-indexing-trap/index.md index ed256e0..f0c3b8d 100644 --- a/blog/content/fpv/betaflight-vtx-indexing-trap/index.md +++ b/blog/content/fpv/betaflight-vtx-indexing-trap/index.md @@ -187,7 +187,7 @@ Reload, and the wheel does what the physical rotation implies: bottom = low, mid Fix the indexing and 400 mW *should* appear at the top of the wheel. If it still doesn't, the VTX itself may be holding it back. Many SmartAudio VTX — the TH3 included — ship with a **restricted (region-locked) mode** that caps output for regulatory compliance. In that state the high-power entries in your `vtxtable` are defined correctly, selected correctly, and simply refused by the hardware. -Restricted mode lives on the VTX, not in Betaflight, so no amount of CLI editing clears it. You disable it on the unit itself — typically a button hold on the VTX (or the matching SmartAudio "unlock") to drop out of the region-locked profile. [TODO: confirm the exact TH3 V02 unlock sequence before publish.] +Restricted mode lives on the VTX, not in Betaflight, so no amount of CLI editing clears it. You disable it on the unit itself — typically a button hold on the VTX (or the matching SmartAudio "unlock") to drop out of the region-locked profile. I have not confirmed the exact unlock sequence on my own TH3 V02, so treat the button-hold as the general pattern rather than a recipe, and check your VTX documentation for the specific one. The tell is diagnostic: if low and mid power select correctly but the top level alone stays dead after the index fix, suspect restriction, not configuration. diff --git a/blog/content/fpv/drone-detector-nn/index.lt.md b/blog/content/fpv/drone-detector-nn/index.lt.md index 3fdd835..5f02f83 100644 --- a/blog/content/fpv/drone-detector-nn/index.lt.md +++ b/blog/content/fpv/drone-detector-nn/index.lt.md @@ -186,10 +186,6 @@ pos_weight = (n_neg / n_pos).clamp(min=NUM_CLASSES, max=30.0) ## Rezultatai ir kiekvienos klasės analizė -*[TODO: Mokymo kreivė — nuostoliai ir makro F1 per ciklus]* - -*[TODO: 10×10 painiavos matrica]* - ### Kiekvienos klasės F1 | Klasė | F1 | Pastaba | diff --git a/blog/content/fpv/drone-detector-nn/index.md b/blog/content/fpv/drone-detector-nn/index.md index 3b893d5..641c58b 100644 --- a/blog/content/fpv/drone-detector-nn/index.md +++ b/blog/content/fpv/drone-detector-nn/index.md @@ -204,12 +204,6 @@ The minimum clamp at `NUM_CLASSES` (10) prevents near-zero weights for majority ## Results and Per-Class Analysis - -*[TODO: Training curve — loss and macro F1 across cycles, phase transitions annotated]* - - -*[TODO: 10×10 confusion matrix]* - ### Per-Class F1 | Class | F1 | Note | diff --git a/blog/content/fpv/meteor75-pro-ii-fighting-resonances/index.lt.md b/blog/content/fpv/meteor75-pro-ii-fighting-resonances/index.lt.md index dd0c8fc..270dda9 100644 --- a/blog/content/fpv/meteor75-pro-ii-fighting-resonances/index.lt.md +++ b/blog/content/fpv/meteor75-pro-ii-fighting-resonances/index.lt.md @@ -19,28 +19,6 @@ series: thumbnail: "meteor75-pro-vs-pro-ii.jpg" --- - - Craft name **Snake**. Pradžioje tai buvo Meteor75 Pro, dabar — Meteor75 Pro II: rėmas ir gaubtas iš AliExpress, viskas, kas kainuoja, perkelta be pakeitimų. Tas pats **Matrix 1S 3-in-1 FC**. Tas pats **narrow-FOV DJI O4** air unit. Naujas kiautas, seni vidūriai, diff --git a/blog/content/fpv/meteor75-pro-ii-fighting-resonances/index.md b/blog/content/fpv/meteor75-pro-ii-fighting-resonances/index.md index 42a6793..0587382 100644 --- a/blog/content/fpv/meteor75-pro-ii-fighting-resonances/index.md +++ b/blog/content/fpv/meteor75-pro-ii-fighting-resonances/index.md @@ -20,44 +20,6 @@ series: thumbnail: "meteor75-pro-vs-pro-ii.jpg" --- - - Craft name **Snake**. It started life as a Meteor75 Pro, and it is now a Meteor75 Pro II — frame and canopy ordered off AliExpress, everything expensive carried straight over. Same **Matrix 1S 3-in-1 FC**. Same **narrow-FOV DJI O4** air unit. New shell, old guts, and by the @@ -2744,7 +2706,6 @@ impact-exclusion pad or the log being different, not a correction. Post-filter r as both 0.50 and 0.67 °/s for the foam because one is a Welch integration above 60 Hz and the other an 80–780 Hz bandpass RMS; they are two metrics, not two answers. - Practices that repeatedly changed the conclusion — not general advice, things that actually flipped an answer in this specific week: