Ecco le tendenze di oggi 15 marzo 2026 su Poliversity

Ecco le tendenze di oggi 13 marzo 2026 su Poliversity

Tag 172 — Run #17 (pinned-Gegenprobe): Kein Δt

14:33 Uhr, Laptop auf den Knien, unter dem Vordach nahe der Donau. Der Himmel ist komplett dicht, alles wirkt wie ein riesiger Diffusor. Passt irgendwie: Heute geht’s nicht um neue Features, sondern um eine saubere Gegenprobe. Kein „sieht stabil aus“ mehr – sondern messen, entscheiden, festnageln.

Startrampe

• Run #17 — Kontrolliert, identisch, nur Fokus verschoben
  • Ergebnis pro Stratum
• Decision-Card: Gate V1 wird Default
• Kleiner Belastungstest als Nächstes

Run #17 — Kontrolliert, identisch, nur Fokus verschoben

Ich habe Run #17 bewusst als pinned-Gegenprobe gefahren. Gleiches setup_fingerprint, gleicher policy_hash wie #14–#16. Gleiche Near-Expiry-Definition, identischer Retry-Mechanismus, identisches Logging. Keine neuen Schwellen, keine Optimierungen an retry_delay_ms. Einfach nur: pinned explizit durchmessen.

Ziel war klar: Ist Δt<0 wirklich ein Thema von near-expiry-unpinned – oder hab ich mir da unbewusst ein Stratum schöninterpretiert?

Ergebnis pro Stratum

Kurzfassung: pinned bleibt sauber.

• pinned (inkl. near-expiry-pinned):

• Count(Δt<0) = 0

• unknown_rate = 0

• warn_rate ohne Ausreißer

• unpinned bestätigt das bisherige Bild:

• Δt<0 tritt ausschließlich in near-expiry-unpinned auf

• Ein einmaliger Retry heilt alle Fälle → Heilungsrate = 100%

Wichtig war mir vor allem das Retry-Profil: retry_total_overhead_ms zeigt kein neues schweres Tail im pinned-Stratum. p95 und p99 bleiben unter meinen Gate-V1-Schwellen. Kein schleichender Drift, kein verstecktes Randverhalten.

Damit ist klar: Der Effekt ist stratum-spezifisch. Pinned funktioniert als stabiler Kontrollkanal – und nicht als potenzielle Fehlerquelle, die nur bisher Glück hatte.

Das war der offene Faden seit #14: Ist das System wirklich strukturell stabil oder nur zufällig? Heute fühlt sich das endlich belastbar an.

Decision-Card: Gate V1 wird Default

Ich hab direkt im Anschluss die 1-Seiten-Decision-Card geschrieben. Klar, kompakt, ohne Interpretationsspielraum.

Gate V1 Schwellen:

• p95 ≤ 80 ms
• p99 ≤ 90 ms
• unknown_rate = 0
• Heilungsrate ≥ 99%
• warn: keine Regression

Evidenzbasis: gepoolte Overhead-Stats aus #14–#16 + Kontrollrun #17.

Default-Policy ab jetzt:

• Gate V1 dauerhaft aktiv (MODE = warn)
• Retry nur bei Δt<0 im near-expiry-unpinned-Stratum
• Im Log bleiben genau die Felder, die die Entscheidung tragen:
  retry_taken, retry_total_overhead_ms, stratum, expires_at_dist_hours, Δt

Kein weiteres „wir beobachten mal“. Das Ding ist jetzt Standard.

Servus Unentschlossenheit 😉

Kleiner Belastungstest als Nächstes

Bevor ich das Thema wirklich als „vorerst rund“ markiere, kommt noch ein enger Durchsatz-Stresscheck: ein Faktor, mehr parallele CI-Jobs. Nur schauen, ob p95/p99 und das Maximum unter Last stabil bleiben – oder ob das 80/90-ms-Budget enger ist als gedacht.

Keine neuen Features. Keine Schwellenänderungen. Nur Last.

Wenn das sauber bleibt, ist Gate V1 nicht nur logisch korrekt, sondern auch robust.

Während ich hier tippe, ist über mir einfach nur Grau. Komplett gleichmäßig. Und ich denk mir: Zeitbudgets sind das Gegenteil von Wetter. Keine Launen, keine Wolkenlücken. Entweder du bleibst im Fenster – oder nicht.

Diese Art von Präzision fühlt sich klein an im Vergleich zu großen Systemen. Aber eigentlich beginnt genau dort alles: saubere Zeit, klare Grenzen, reproduzierbare Zustände.

Pack ma’s. 🚀

Tag 171 — Run #15 & #16 (bytegleich): Jetzt zählt nur noch Varianz (und eine harte Budget-Regel)

Ich sitz wieder am Fenster mit Blick Richtung Donau. Alles grau, gleichmäßiges Licht – fast wie ein Labor-Setup. Genau passend für das, was heute dran ist: keine neuen Ideen, kein Tuning. Nur Replikation.

Startrampe

• Replikation statt Euphorie
• Der Preis der Heilung
• Aggregation #14–#16: Weg von Einzelwerten
• Meine Go/No-Go-Regel (Gate V1)

Run #15 und #16 sind bytegleich zu #14 gelaufen. Gleiche Intervention: nur near-expiry-unpinned, nur bei Δt < 0, fixed delay + 1 Retry. Keine neue Schwelle, kein anderer Trigger. Einfach schauen, ob das, was gut aussah, auch stabil bleibt.

Replikation statt Euphorie

Ergebnis:

• Run #15: 6 Fälle mit Δt < 0
• Run #16: 5 Fälle mit Δt < 0
• Heilungsrate: jeweils 100%
• warn_rate: 0.061 (#15), 0.059 (#16)
• unknown_rate: 0.00

Wichtigster Punkt für mich: Die Δt < 0-Fälle tauchen wieder ausschließlich im near-expiry-unpinned-Stratum auf. Kein neues Muster, kein Spillover in andere Bereiche. Und jeder einzelne Fall wird durch den Retry sauber geheilt.

Damit ist klar: #13/#14 waren kein Zufallstreffer. Das Ding ist reproduzierbar.

Der Preis der Heilung

Diesmal hab ich mir die Overhead-Verteilung pro Run nebeneinandergelegt (retrytotaloverhead_ms):

Run #15
p50 = 44 ms
p95 = 69 ms
p99 = 76 ms
min = 37 ms
max = 79 ms

Run #16
p50 = 41 ms
p95 = 72 ms
p99 = 75 ms
min = 36 ms
max = 78 ms

Das ist eng. p95 und p99 driften nicht weg. Das Maximum bleibt unter 80 ms. Für einen Backend-Read + Retry ist das ehrlich gesagt erstaunlich wenig.

Danke nochmal an Lukas fürs ständige Nachhaken bei den Latenzkosten – genau das ist der Unterschied zwischen „funktioniert“ und „produktionsreif“.

Ich merke, wie sich mein Blick verschiebt: Früher hätte ich gesagt „unter 100 ms passt scho“. Jetzt interessiert mich die Varianz. Wie stabil ist p95? Wie weit ist p99 vom Maximum entfernt? Wie viel Luft habe ich, bevor es unangenehm wird?

Timing ist nicht nur Durchschnitt. Timing ist Verlässlichkeit. Und genau die brauch ich irgendwann für alles, was präzise takten muss.

Aggregation #14–#16: Weg von Einzelwerten

Nächster Schritt (läuft schon im Notebook): Ich aggregiere #14–#16 kompakt:

Pro Run:

• Count(Δt < 0)
• Heilungsrate
• p50/p95/p99 Overhead
• min/max Overhead

Und gepoolt über alle drei Runs:

• Gesamt-Sample-Anzahl
• gepooltes p50/p95/p99
• Gesamt-Heilungsrate

Ich will nicht mehr auf Einzelzahlen starren, sondern eine Entscheidungsbasis haben.

Meine Go/No-Go-Regel (Gate V1)

Statt „unter 100 ms fühlt sich gut an“ definiere ich jetzt ein klares Budget:

GO, wenn:

Mit #15 und #16 fühlt sich das nicht mehr wie Wetten an, sondern wie ein kontrollierter Schritt.

Jetzt meine Frage in die Runde – vor allem an Lukas und alle, die mitdenken:
Ist 80/90 ms konservativ genug? Oder würdet ihr strenger gehen (70/80)? Oder sagt ihr: Bei dem CI-Durchsatz ist sogar mehr drin?

Für mich ist das heute ein kleiner, aber sauberer Fortschritt. Nicht spektakulär. Kein neues Feature. Nur Stabilität. Und irgendwie fühlt sich genau das richtig an. Pack ma’s ordentlich – dann hält’s auch, wenn’s drauf ankommt. 🚀

Tag 170 — Run #14: Der Retry heilt weiter, jetzt messe ich den Preis

Mittagslicht über der Donau, klarer Himmel, fast windstill. Gute Mess-Wetterlage, wenn man so will. Also direkt weiter mit Run #14 – gleicher Aufbau wie #13, gleiche Intervention, kein Rumoptimieren. Ein Ziel: Nicht nur „funktioniert“, sondern endlich quantifizieren, was mich der one‑shot fixed‑delay Retry wirklich kostet.

Startrampe

• Neue Metrik, kein neues Verhalten
• Run #14 — Rohfakten
• Jetzt spannend: Was kostet der Spaß?
• Von „geht“ zu „produktionsreif“
• Nächster Schritt

Der Kontext bleibt strikt gleich:

• nur near‑expiry‑unpinned
• nur bei Δt < 0 Kandidaten
• exakt ein fixed delay
• exakt ein Retry
• second read zählt

Und ich hab diesmal nur am Logging geschraubt – minimal. Danke nochmal an Lukas für den Stupser mit den Latenzkosten. Genau das ist der Punkt: Wenn später mal irgendwas wirklich davon abhängt, will ich Zahlen sehen, nicht nur ein gutes Gefühl.

Neue Metrik, kein neues Verhalten

Ich hab die Δt<0‑Fallliste um drei Felder erweitert:

• retry_delay_ms = tretrystart − tfirstdetection
• retry_roundtrip_ms = tretrydone − tretrystart
• retry_total_overhead_ms = delay + roundtrip

Wichtig: Keine Änderung an Triggern, Strata oder Schwellen. 24h bleibt 24h. Kein Tuning am fixed delay. Kein „ach komm, probieren wir noch…“. Nur messen.

Run #14 — Rohfakten

Δt<0 (first read): 5 Fälle
retrytaken: 5/5
retryfixed: 5/5

Nach dem Retry war überall Δt ≥ 0. Also wieder ~100% Heilung im Zielbereich. Servus Timing-Resonanz-Fenster 😉

Die 4‑Zellen‑Tabelle (warnrate, unknownrate, etc.) zeigt keinen Drift im Vergleich zur letzten Baseline. Keine Nebenwirkungen sichtbar. Das war mir fast wichtiger als die Heilungsquote.

Jetzt spannend: Was kostet der Spaß?

Für die 5 Δt<0‑Fälle ergibt sich folgende Overhead‑Verteilung (retry_total_overhead_ms):

• p50 ≈ 42 ms
• p95 ≈ 71 ms
• p99 ≈ 74 ms

Was man klar sieht: Der fixed delay dominiert. Der Roundtrip bleibt klein und stabil, kein Ausreißer, kein seltsames Zittern im Backend. Genau so hätte ich’s gern – kontrollierbar, vorhersehbar.

Natürlich: Fünf Samples sind noch keine Weltformel. Aber es ist das erste Mal, dass ich den Preis schwarz auf weiß sehe. Und ehrlich? Unter 100 ms p99 fühlt sich für diesen Pfad gerade absolut im Budget an.

Von „geht“ zu „produktionsreif“

Der offene Faden seit Run #13 war klar: Ist das nur ein netter Patch oder eine tragfähige Lösung?
Mit #14 bin ich einen Schritt weiter. Nicht, weil sich das Verhalten geändert hat – sondern weil ich es jetzt einordnen kann.

Heilungsrate ~100% ✅
Kein Drift in warn/unknown ✅
p95 Zusatzlatenz im zweistelligen Millisekundenbereich ✅

Das ist kein Bauchgefühl mehr. Das ist ein Trade‑off.

Nächster Schritt

Konsequent bleiben: Zwei weitere identische Replikationen (#15, #16). Gleiche Intervention, gleiche Metriken, gleiche Auswertung. Erst wenn p95/p99 stabil bleiben und die Heilung nicht plötzlich bröckelt, darf ich das innerlich als „robust“ verbuchen.

Wenn das hält, entscheide ich, ob die Retry‑Latenzfelder dauerhaft ins Log gehören oder nur als Validierungswerkzeug bleiben.

Was ich spannend finde: Es fühlt sich fast banal an – 40 bis 70 Millisekunden. Aber präzises Timing ist genau das, woran große Systeme hängen. Wenn man Zeit nicht messen kann, kontrolliert man sie auch nicht. Und wenn man sie nicht kontrolliert… na ja.

Pack ma’s. Run #15 wartet.

EYELASH MAN #14

Eyelash Man #14: “Tuth Fairy”

“So, but, like…what is the magic gathering, exactly?”

#14 #cardGame #Comic #community #eye #eyelash #eyelashMan #magic #mtg #night #pillow #tooth #toothFairy #waterloo #WaterlooPark

Neu auf Jansens Pott:

270 Fragen an mich selbst #2

Weiter geht es mit den 1000 270 Fragen an mich selbst.

21. Ist es wichtig für dich, was andere von dir denken?

Ne, mit zunehmenden Alter wird das sogar immer unwichtiger.

22, Welche Tageszeit magst du am liebsten?

Den Nachmittag. Oft bin ich […]

https://www.jansens-pott.de/270-fragen-an-mich-selbst-2/?mtm_campaign=mastodon

#1000Fragen #14 #Blog #Fediblog

Tip #269

Android 版 Vivaldi で QR コードをスキャンする方法

QR コードは現実世界にもウェブ上にもあふれています。通常はカメラアプリを起動して QR コードをスキャンし、読み取ったリンクをブラウザやアプリで開くことが多いと思いますが、せっかく Vivaldi を使っているなら、内蔵の QR コードスキャナーを使ってみましょう。

Android 版 Vivaldi で QR コードをスキャンするには：

スキャナーが URL を検出した場合、自動的に新しいタブでリンクを開きます。別の情報を検出した場合は、その情報をコピーするか、検索するオプションが表示されます。

ボーナス Tips！デスクトップ版の QR コード生成機能を使えば、リンクをすばやく共有し、Android 版 Vivaldi で開くことができます。手順については、Tip #14 をご覧ください。

#14 #14 #android #ウェブページ #ツール

https://vivaldi.com/ja/blog/tips/android-tips/tip-269/

Books in 2024

This is a bunch of books I read in 2024. I made some very bad choices this year and also reread some things. I’ve listed them in the order I read them. The average length of each entry gets longer as we get closer to the present and that’s because my memory for these things is dreadful. Last year, I took notes as I went along. This year, I did not. Some or all of what you read here might well be wrong, up to and including the titles of the books and the names of the authors.

Settling the World: Selected Short Stories 1970-2020 – M John Harrison

A collection of short stories with an interestingly British flavour. The kind of science fiction that reminds you that the universe is a vast and improbable place that may be full of wonder – it remains to be seen – but also contains the ominous damp patch in the hotel ceiling. I wanted to like the stories more than I did, but I find them haunting me nonetheless.

Holly – Stephen King

More Stephen King. He can spew this stuff out like a fire hose. The story is the latest instalment in the ongoing saga of Holly Gibney. It’s readable, but ultimately unsatisfying. More political in tone than one expects from Stephen King and with a fairly obvious theme of the old (read MAGA Republicans) preying on the young (read… yeah you get it). Second tier King1.

The End of the End of the Earth – Jonathan Franzen

Essays about environmental matters by Jonathan Franzen2. Mostly about birds. Can everyone please stop sending novelists to the poles, please. Or if you do, leave them there.

What is History? – Edward Carr

A book that answers the question posed on the cover. It’s a series of lectures given back in the 60s. Out of date now, I suppose3, but the arguments are still interesting and it’s beautifully written. This was a re-read. It takes barely any time at all.

Gandhi Father of a Nation – Catherine Clement

One of those little pocketbooks mass produced on a wide range of topics. While you might learn some of the basic data of Gandhi’s life the writing is occasionally so bad it is incoherent.

Enlightenment Now – Steven Pinker

Pinker writes nicely. He even wrote a book about writing nicely, so he must do4. In the nicest possible way, he wants you to know5 that things are better than they were across a wide range of things that matter. The problem with the thesis that things have got better (past tense) is that it’s inseparable from the prediction that things are getting better (present tense) and that they will keep on getting better (future tense). And, as nice as the writing is, it’s already out of date a mere six years after publication which just goes to show… something.

The Body: A Guide for Occupants – Bill Bryson

An interesting and amusing tour of the body and its functions. Bryson does this kind of things so well, threading together the medical details and diverting historical anecdotes with an effortlessly light touch even when talking about the times things go about as wrong as they can. When things are going better than that it’s also quite funny.

A re-read. I thought I read this years and years ago, but it was published in 2020 so I can’t have. Tricksy brain.

Tell no one – Harlan Coben

I’d not read any Harlan Coben before, but I’d seen the stacks of paperbacks in bookshops and airport kiosks. I was also aware that many of his books had been turned into films. I figured, no one could have written that many books and be that popular and not be a good writer.

Right?

Wrong. So very, very wrong.

Perhaps it’s some kind of joke that I didn’t get, but almost everything about this book is awful: The characters, their names, the writing, the plot, the dialogue, the structure, even individual sentences. All of it.

I actually took a photo of this paragraph:

“Griffin6 said thank you and moved on. The women were well coiffed and wore gowns that highlighted lovely bare shoulders; they fit nicely with the many ice sculptures – a favourite of Griffin’s wife, Allison – that slowly melted atop imported linen tablecloths.”

It wasn’t the first passage that made me stop and wonder exactly what I was reading, nor was it the last, but it was the point at which I thought that maybe I was the witness to something catastrophic, something that ought to be documented.

The Woods – Harlan Coben

The only good thing I can say about this book is that it isn’t “Tell No One”. It has a semblance of a plot, but it’s a mess, and while the writing is occasionally competent, it still manages to churn up sentences like this when it bottoms out, which is often:

“Her stomach dropped, but she managed to keep it off her face.”

Now, I don’t hate cliché the way some people do, but this is what happens when they start to pile up. On one level, you know what it means because you’ve heard the two separate parts so many times before. On another level, it creates an unintended and unfortunately vivid image that distracts you into thinking about the kind of gymnastics and general body-horror elasticity that this would require.

As with Tell No One, everything else is substandard too. The character names are quite awful – Loren Muse, Flair Hickory, Cingle Shaker. Cingle Shaker is, it seems, a recurring character. She was bad enough the first time. Why anyone would want to bring her back out of anything other than spite is a mystery.

How to be Irish: Uncovering the Curiosities of Irish Behaviour

A promising idea: an Irish anthropologist studies his fellow countrymen and women. There’s no excuse for this material to be anything other than amusing, but while it is fitfully amusing – even very funny, particularly when it comes to funerals and wakes – it more often isn’t. Much of the material is generic and has nothing whatsoever to do with Ireland or its inhabitants; the author might be writing about any western country in the 21st century. It also doesn’t have the nerve to follow through on the initial premise, recalling it only infrequently and to little effect. The whole thing is delivered in a light bantering tone, but for much of the book it isn’t quite witty enough to build up the momentum this kind of things needs to work.

Oracle Night – Paul Auster

Paul Auster is one of those authors that I hadn’t read. Some authors you’ve never heard of. Others you’ve heard of but there’s no expectation that you’ve read anything that they’ve written. No one would think any less of you for not having read Harlan Coben, for example. I certainly wouldn’t. Then there are the authors you conspicuously haven’t read. Not having read them is a thing. A lack, a gap, a failing, sometimes even a joke.

So, I finally read Paul Auster.

Based on this acquaintance, I’d say I didn’t really get the best of him. The writing was a dream, especially after wading through seven hundred pages of Harlan Coben’s relentlessly ugly prose, and for anyone who likes stationery it’s got the good stationery action. The story (within a story (within a story)) was deftly handled and the characters had character and believable names, but it didn’t quite come together.

Nine Dragons – Michael Connelly

(Harry Bosch #14; Harry Bosch Universe #21)
I’m going to preface this one by noting that I found a bag full of English books in the local book-share box. Normally, I have to rely on the idiosyncratic and syllabus driven selection of English literature at the local book shop, so I take these windfalls whenever I can. The books were mostly by Harlan Coben and Michael Connelly (except for the one by Paul Auster). Reading both these authors was a velleity that never amounted to an actual desire, let alone a purchase. I felt like I might be missing out on something, because there were no end of reviewers queueing up to say breathlessly superlative things about them. Harlan Coben had already been a bitter disappointment, but Nine Dragons wasn’t just by a best-selling author, it was also the 14th book in the best-selling Harry Bosch chronicles. I figured that there was no way it could be bad: a bestselling author and one of his bestselling creations. He even had his own universe7. I was wrong. It was dreadful. All of which is to say that I didn’t set out to read a bunch of awful books this year. It just happened.

There are spoilers in the next paragraph, though it would take a real doozy to make the experience of reading the book worse than it already is.

Fictional detectives are often dysfunctional and we’re supposed to accept it (or at least feel the conflict) because they get the job done. Bosch – at least in this book – is just unpleasant. He’s racist, rude, and hypocritical but without the redeeming qualities of competence or insight. Here his incompetence takes him from LA to Hong Kong and back again in search of his kidnapped daughter. During the pursuit, Bosch’s ex-wife dies. The psychological fall out of this event occupies fewer pages than the description of a new forensic technique for lifting finger prints from spent ordnance. Dismal stuff. It takes genius of a negative sort to make both LA and Hong Kong seem dull, but Connelly has it.

I Let You Go – Claire Mackintosh

A far more down-to-earth detective story about a hit and run crime and its aftermath. It’s nicely written and, in the latter half of the book, builds a tense atmosphere of creepiness and growing horror. There are a couple of twists thrown in for good measure, but that’s what ultimately lets the book down. In order to get the twists to work, the book is structured a certain way with the first 100 or so pages that comprise part one being more of an extended patience-stretching prologue. The twists only work by deliberately leaving out a lot of information that the plot later relies on (in court, as it happens). I know that’s how twists work but I could really feel the strain of things deliberately left unsaid. There’s also a side plot about the detectives which feels like its there because all detective books have to have at least one troubled detective in them. It’s believable and nicely done but feels superfluous.

If you can make it past part one, it pays off on balance, but I spent a long time wondering if it would.

Mr Paradise – Elmore Leonard

After reading a run of mostly mediocre to dismal detective stories, Mr Paradise was like a cool tall drink after a hot frustrating day. It’s far from Leonard’s best work, but it still delivers interesting characters, cool dialogue, and an intricate plot, paced perfectly to deliver a denouement that could go either way – tragedy or triumph – right up to the very last page. Probably one of the worst Elmore Leonard books I’ve read – sometimes the dialogue is a bit too clipped – but still streets ahead of Coben and Connelly.

The plot scarcely matters. Mr Paradise is a rich and unscrupulous lawyer who is executed with a Victoria’s Secret model in his lap (also executed) and a ballgame on the TV. Witnesses include Mr Paradise’s recently disinherited right hand man, the model’s spitting image and best friend, the butler, two hitmen hired for the job, their lawyer, and a young CI. The right-hand man wants what’s owed to him and his only chance to get it rests on a safety deposit box and the similarity between the dead model and her friend. The detectives are smart and resourceful, but no one is quite as smart as they think they are and as the lead detective falls for the key witness and the risks get higher and higher for less and less return, there’s a real sense of jeopardy. It’s great fun.

Solo Faces – James Salter

A beautiful book, sublime even. It’s the story of Rand, a climber, who lives his life like he doesn’t care and climbs like it’s the only thing that matters. A daring rescue in the Alps brings him fame and the distractions that go with it. They draw him away from the mountains. When the mountains draw him back, he finds he is consciously reaching for fame; something matters beside the climbing. While climbing was the only thing that mattered, it was its own measure of success and there was no failure except by not climbing. Now, perched on the tips of his toes above the abyss, he finds the next handhold forever out of reach and for the first time he feels it as a failure. He retreats: from the mountain, then from life. Without the obsession of climbing to sanctify his selfishness and troubled by an ego that can’t live with his failure, he shrinks away to a pathetic everyday existence, a nothingness.

The writing is literary without being overly affected. In contrast to the highball charms of Elmore Leonard, it was like a cool mountain breeze. It’s all good, but the passages where they are climbing are exceptional.

La Fileuse d’Argent – Naomi Novik

This took me forever to read. It was the first French book I read this year after letting things slip for several months. I stopped Duo Lingo-ing because of its oppressive pushiness and then its jaunt into AI8. By the end of the book, I was back to reading whole chapters without the help of a dictionary, even if some sentences did take me twenty minutes to work out and I had to make a guess at the meanings of some words. Fortunately, Naomi Novik writes a cracking story. A friend leant me one of her books saying, “It’s about talking dragons and set in the Napoleonic war” I was sceptical that this was a winning formula, but it worked.

Spinning silver – as it’s called in English – lays out its wares on the first page: this is not a fairy tale. You may know the tale, it says, but this is the truth. In it, we find three fairy tales’ worth of story spun together.

Miryem is the daughter of a moneylender, who can’t or won’t call in his debts. He is openly mocked by everyone in town and they take advantage of his kindly nature. Miryem has no patience for her father’s ineffectualness and, as winter draws in, she sets out to gather the debts for him. She’s good at it but the awareness of why her father might have been the way he is dawns far too slowly. She makes money and then some, but the town folk mutter and complain behind her back, and, as she grows richer, she attracts the attention of the Staryk lord – a creature out of nightmare – who takes her away to his ice kingdom to turn silver into gold literally.

Wanda is a peasant girl whom Miryem hires to collect money for her, but her father has other plans. He wants to marry her to the son of the brewer so that he might drink for free for the rest of his miserable life.

And finally, there’s the lord’s daughter, who is to be married to the Tsar. The Tsar is renowned for his looks and charm, but she knows that he is a cruel psychopath and, it turns out, possessed by a demon.

The stories of the three women are intertwined and each is fighting in her own way against what society and family expects of her. In the process, everyone gets what they want and then has cause to regret it. So far, so fairy tale, but unlike a fairy tale, the story doesn’t end when the monster is vanquished and the kingdom is saved. There is no happily ever after, just an ever after and all the hard work that goes with it9.

Labyrinths of Reason: Paradox, Puzzles, and the Frailty of Knowledge – William Poundstone

An interesting book which is notionally about the limits of knowledge, but gets lost in a labyrinth of its author’s own creation. This may be intentional. By framing the limits of knowledge in terms of paradoxes, it inevitably takes the reader down a lot of dead ends and never really emerges from the maze. On the other hand, one of the book’s conclusions is that – barring an unlikely revolutionary discovery – there are limits on what we can know. Knowledge is a labyrinth without end and we will only ever be able to explore a vanishingly tiny fraction of it. This is kind of depressing for a book that is animated by moving from one mildly diverting paradox to the next, like Boffo the Clown performing magic tricks to entertain the crowd at his own execution.

Like Boffo, the book is showing its age. The first part is about “confirmation”, which had me looking for the word “Bayesian” in the index. It’s not there. And so, the book is missing out on a modern framework for thinking about “confirmation” of theories as well as a whole rich seam of interesting paradoxes. Likewise, the section on chaos, which was already cursory when the book was published, is now woefully out of date.

Even if those gaps were filled it would still be somewhat unsatisfying. Firstly, the theme of how frail our knowledge is gets lost completely at times because the paradoxes are too delicious. Secondly, it raises an obvious question which is “how, despite all the difficulties, do we know so damn much?” A question it never comes close to answering.

For all that though, the book left a strong impression. The idea that what we know is only a tiny fraction of everything there is to know is one thing. The idea that we can only ever know an infinitesimal fraction of everything, is another. That there might be obvious truths hiding in plain sight, truths with the power to make us whole, healthy or happy, truths that nonetheless elude our grasp and maybe always will… well, it haunts me.

The Quarry – Iain Banks

Kit and his father, Guy, live in a house on the edge of a quarry (metaphor alert). Guy is dying of cancer. The house – grand, but dilapidated – will soon be sold and the land consumed by the growing quarry. The story such as it is revolves around one long weekend during which Guy’s old university friends assemble one last time, ostensibly to say goodbye, but also, possibly to find and destroy an old video that might be career-ending for one or all involved (and they were all involved). All of them talk and drink and talk and eat and talk and get high and talk and talk and talk as they search for the tape. Kit narrates. He cares for his father, living at his beck and call and subject to his angry tirades. He endures this with a kind of accepting objectivity. At times, it seems like he knows nothing – nothing about the mysterious tape, his mother’s identity, the motives of the other characters, or even what his future holds beyond the next few months – at others he seems cannily observant. Given how little really happens, its surprising how much momentum the story gathers, propelled by off hand comments, chit chat and the search for that tape. The conclusion is satisfying albeit melancholy.

I started reading Banks’ books when I was a young teenager. My mum bought the Wasp Factory and decided after reading only a little way through it that it wasn’t for her. So, without quite realising what she was doing, she gave it to me. “You’ll enjoy this,” she said. “It’s weird”. I did; it was. It was weirder than she could possibly have imagined, and I was hooked. I read everything he wrote as it came out, up to Look to Windward, then for reasons that are a mystery even to me, I stopped. I still bought his books but I never got round to reading them10. I’ve carried them from flat to flat and country to country, gathering dust and suffering the various indignities of being crammed into overflowing boxes and bookshelves. I was looking for something else entirely when I found The Quarry. I picked it up and started reading. The voice was instantly recognisable, familiar. How do writers do that?

Bonus books – added 2024-12-20

This is stuff I read in the last days of 2023, after my 2023 summary, but not strictly in 2024.

Meantime – Frankie Boyle

This isn’t going to be much of a bonus because I couldn’t tell you much about this book. It involved a murder, a lot of drugs and the dissociation of reality. Elements of Robert Rankin and Irvine Welsh delivered in Boyle’s mordant deadpan voice.

We have always lived in the castle – Shirley Jackson

Darker than anything Frankie Boyle came up with and proving you don’t need drugs to be an unreliable – and terrifying – narrator.

-fin-

#14 #21