Five and a half years of monthly newsletters, examined in glorious nerdy detail.
Dann's monthly word count has grown 61% from the first issue to the latest year. The format stabilized at 7 sections per issue in January 2021 — but Dann started filling those sections with more.
Each point is one issue. Trend line is a 6-issue rolling average.
2026 represents only the first five issues of the year (Jan–May).
From October 2020 through May 2026, Dann has published — issues across — possible months. April 2021 is the only month that ever went unpublished — a rare and lonely gap in an otherwise relentless monthly rhythm.
Each tile is one month. Darker = longer issue. Hover for details.
Themes drift in and out of the newsletter across years. The story of 2022–2026 is the story of AI eating everything — and crypto evaporating just as it arrived.
Counts the number of times keywords for each theme appear in the body. Toggle themes in the legend.
Mentions of AI/LLM/Claude/OpenAI/etc. per issue. The curve speaks for itself.
After stripping URLs out of the link soup, the real word-count winners emerge. AI is mentioned 198 times across the archive (the genie). Apple comes second at 109, then Google, YouTube, TikTok, and iPhone. Counts below exclude URL fragments — only narrative mentions.
Try "crypto" to see the bubble pop, or "Anthropic" to watch it appear out of nowhere in 2024.
The records, the extremes, the outliers. Things you do not realize you have done across 67 issues.
Emojis are core to The Dann Chronicles' visual identity. Most issue title emojis are used only once across the whole archive — a deliberate non-repeating set, like fingerprints for each month.
Surfaces the recurring "themed segments" — 🤖 robots, 📺 TV reviews, 🪦 dead products, 📱 tech.
— title emojis appear exactly once. Repeats are noted with a small badge.
— words appear exactly once across all 67 issues. After filtering out plurals, junk, and very common English, here are the most interesting one-hit wonders — categorized.
Uncommon English vocabulary you reached for exactly once. Hover to see which issue.
Words that don't appear in standard English corpora — likely your inventions, fusions, or playful mashups.
People, products, places, and brands you named exactly once.
Words that appear once because they're misspellings. Light copy-edit candidates for the future.
— words land in the rare bucket — used a few times, but never enough to feel routine. The best of these reveal recurring obsessions and the occasional invented vocabulary that almost caught on.
Vocabulary you reached for more than once but kept rare — often the most distinctive choices.
Names and unusual words that made multiple appearances without becoming a fixation.
Across 84,000-ish words, Dann uses around 9,000 unique words. New vocabulary keeps trickling in even on issue #67 — the curve never quite flattens.
Each issue still adds 50–200 brand-new words to the lifetime vocabulary.
You started using AI to proofread the newsletter sometime in 2023. Did it actually move the needle on typo rates? Here's what the archive shows.
The big drop happened before AI editing — between 2021 and 2022 — when you apparently got more careful on your own. AI editing then continued the decline through 2024–2025.
—% drop in typo rate from your worst year (2021) to your best (2025).
Detected by cross-referencing each word against a large English corpus and looking for edit-distance-1 variants that are dramatically more common. Manually filtered to exclude proper nouns and intentional creative spellings.
"This this", "but but", "was was" — easy for a spell-checker to miss but obvious to an AI proofreader.
The typo rate dropped ~80% between 2021 and 2022 — before you started using AI for editing — which suggests you were already getting more deliberate as the newsletter matured. From 2023 onward, AI catches the long tail: the kind of one-off typos a human writer trained on their own draft has trouble spotting in re-reads. The average sentence length stayed remarkably steady throughout (16–18 words), which lines up with your note that AI only proofs but doesn't restructure prose.
Dann is a question-asker, not an exclaimer. The newsletter has 410 question marks to just 59 exclamation points. Em-dashes have surged in 2025 — apparently it was an em-dash year.
Some months consistently produce longer issues than others. The December issue is statistically Dann's longest each year — perhaps because there's more to wrap up.
Following the data-scientist and editor critique, here's a second pass with the higher-rigor and editorial-craft pieces that the first version was missing.
Asked rigorously rather than asserted: where in the time series does the typo rate actually break? Ran a PELT change-point detector on the per-issue rate.
The single most powerful breakpoint sits at August 2021 (ed#10), not 2023 when AI editing entered the picture. That early break is when your typo rate dropped from ~1 per 1k words to ~0.1. The 2023 AI-editing era looks like a continuation of that improved baseline, not the inflection point that caused it. Both t-tests are statistically significant (p < 0.05), but the 2022 break has the bigger effect size by far.
1,065 outbound links across 67 issues, pointing to — unique domains. Here's where you actually send readers.
YouTube has been your #1 nearly every year. Your own blog (dannb.org) is right behind. Amazon affiliate links surged to the #1 slot in 2025.
— section openings extracted. Median length is — characters. — open with a question; —% start with "I". Click below to browse the full gallery.
Beyond the "So long, and thanks for all the [X]" gag, every emoji is essentially a recurring beat with its own DNA. Here's what each emoji-tagged segment looks like across the run.
People who appear in 4+ issues. First-name / last-name variants are merged ("Musk" and "Elon" are folded into "Elon Musk," "Altman" into "Sam Altman," "Bobby" into "Bobby Fingers"). Names that turned out to be products (Ghost CMS, Binding of Isaac) or interjections ("Man,") are excluded.
How accessible is the prose, and how has the rhythm changed?
A modest drift upward: from 5.4 (about 5th-grade reading level) in 2020 to 6.8 today. Still very accessible — the newsletter has never crossed into "you need to concentrate" territory.
Paragraphs per issue have grown from ~23 to ~42, while average paragraph length has gone slightly shorter (42 → 38 words). More paragraphs, slightly snappier each — a more punchy rhythm.
For each year, the words that appear unusually often that year compared to all other years. A vocabulary fingerprint.
2020 was COVID/Chappelle. 2021 was crypto. 2022 was Musk/Twitter. 2024 was Balatro + Macaulay Culkin. 2025 was Burning Man + denim. 2026 has been Claude + Anthropic. Each year has a clear vocabulary signature.
The "December is longest, July is shortest" claim, tested with bootstrapped 90% confidence intervals (2,000 resamples per month).
The "July is your shortest month" claim survives. July's upper bound (1,143 words) sits below most other months' lower bounds. The "December is your longest" claim does NOT survive — December's CI (1,199–1,485) overlaps massively with August, September, October, November, April, and February. Calling any specific month "your longest" was numerically irresponsible. July as a clear summer slump, on the other hand, is real.
Cosine similarity between every pair of issues using TF-IDF vectors. Surfaces unusually similar pairs (perhaps accidental repeats?) and the issues that are most unlike anything else.
All pair similarities are under 0.34 — your issues are remarkably independent of each other. No accidental near-duplicates.
These issues stand out the most from your usual pattern.
For each issue, the most interesting words that appeared in The Dann Chronicles for the very first time. Read in order, it's a six-year vocabulary diary.
Sortable table of every issue. Themes are inferred from body keyword matches.
| Ed | Date | Emoji | Words | Min | Links | Top themes | Summary |
|---|