| name | linkedin-post |
| description | Write a LinkedIn post for one of Dennis's blog articles, in his proven voice. Use when the user says "write a LinkedIn post for <article>", "do a LinkedIn post like the last one", references the Datadog/observability post style, or asks to promote a published blog post on LinkedIn. Includes image suggestions. |
Write a LinkedIn post in Dennis's voice
Turn a published blog article into a LinkedIn post that reads like Dennis wrote it, not like marketing. Two posts have worked well and are the gold references — match their structure and voice, don't invent a new one.
Output is the post text plus image suggestions. End by running the post through the humanizer skill's checklist (see "Final pass").
The formula (reverse-engineered from the posts that landed)
Follow this arc. Keep paragraphs short, rhythm varied, language plain.
- Concrete hook + one fixation. Open on a specific, factual scene (a dated event, a phrase everyone repeats). Then narrow to the single word or idea you got stuck on, set in bold Unicode (Mathematical Bold, the same glyphs as the examples: 𝐒𝐭𝐨𝐜𝐡𝐚𝐬𝐭𝐢𝐜, 𝐍𝐨 𝐨𝐧𝐞 𝐝𝐞𝐬𝐢𝐠𝐧𝐞𝐝 𝐢𝐭). Patterns that work: "Most people read it for X. I got stuck on one word: 𝐘." / "The honest answer: 𝐘."
- Plain-prose framing. Set up the old assumption or the history. "For 20 years..." / "No committee ever decided...". This is the setup the punch lands against.
- The break or the non-obvious fact. What changed, or the thing nobody noticed. Punchy, factual.
- The reframe. State the subject's bet (or the twist) — often a second bold-Unicode phrase (𝐅𝐫𝐨𝐦 𝐌𝐨𝐧𝐢𝐭𝐨𝐫𝐢𝐧𝐠 𝐭𝐨 𝐎𝐛𝐬𝐞𝐫𝐯𝐚𝐛𝐢𝐥𝐢𝐭𝐲 𝐭𝐨 𝐀𝐮𝐭𝐨𝐧𝐨𝐦𝐲 / 𝐨𝐩𝐩𝐨𝐬𝐢𝐭𝐞). Cap it at two bold phrases per post — the fixation and the bet/twist.
- The pivot line: a short standalone line, "Here's the part that stuck with me." / "That's the part that stuck with me." This introduces the share-worthy nugget.
- Honest caveat. Name the weakness. "Some of it is still slideware, and I poke at that in the post." / "This is Part 1 — I haven't gotten to X yet." No overselling.
- The universal turn. Make it bigger than the company/topic. "It isn't really Datadog's question. Anyone building infra runs into it sooner or later." / "Worth remembering the next time someone tells you the current setup is just 'how it's done.'"
Full read: <url> — https://blog.fnil.net/<slug>/. (Dennis runs it through LinkedIn's own shortener.)- Five topic hashtags, relevant to the piece (e.g.
#Observability #SRE #Monitoring #DistributedTracing #DevOps). Always include #Observability if it fits.
Voice rules
- First person, with opinions and mixed feelings. "I got stuck on", "the part that stuck with me", "I poke at that".
- Plain and direct. No corporate buzzwords, no "exciting times", no slogan endings.
- Straight quotes
" only, never curly. No emojis. The only decoration is the 1–2 bold-Unicode emphases.
- Intellectual honesty is the differentiator — the caveat (step 6) is not optional. It's what makes readers trust the take.
- The closing must generalize: the reader should feel "this is my problem too," not "nice writeup of someone else's slides."
Final pass (humanizer)
Run the draft against the humanizer skill before delivering. Watch especially for:
- Rule-of-three overload. Some triads are fine (especially if the topic is three of something), but don't stack three separate triads in a row.
- Duplicate phrasings. Don't use "stuck with me" and "stays with me" in the same post; pick one.
- Self-labeled takeaways. Cut "The takeaway that stays with me:" — just state the takeaway.
- Round-number / corporate softeners. "nearly a decade", "any incentive to" → tighten.
Keep the bold-Unicode emphasis through the humanizer pass — it's Dennis's deliberate LinkedIn voice, not an AI tell, even though humanizer flags boldface.
Image suggestions (always include)
- Lead with the blog cover as a native upload, not a link preview. LinkedIn throttles posts with external links; native images get more reach, and the cover already carries the title + hook. Put the link in the body (proven to work) or in the first comment if throttling is a concern.
- For more engagement, a 2-image carousel: image 1 = cover (the hook); image 2 = a "reveal" image (a diagram from the post) with a one-line caption that detonates the twist, e.g. "The man who drew this in 2017 proposed scrapping it in 2018. Nobody noticed." The payoff lands when the reader swipes.
- Don't make a low-res or academic diagram the single lead image — no title, no hook, weak cold open.
Gold reference posts
Match the cadence of these. They are the standard.
Datadog post (the one that drew a lot of attention):
Datadog dropped a 200-page Investor Day deck in February. Most people read it for the revenue curves. I got stuck on one word: 𝐒𝐭𝐨𝐜𝐡𝐚𝐬𝐭𝐢𝐜.
For 20 years, observability fought complexity. [...] But the system underneath stayed deterministic: same input, same output. [...]
AI breaks the assumption. Same prompt, different answer. [...]
Datadog's answer is to drag the whole field 𝐅𝐫𝐨𝐦 𝐌𝐨𝐧𝐢𝐭𝐨𝐫𝐢𝐧𝐠 𝐭𝐨 𝐎𝐛𝐬𝐞𝐫𝐯𝐚𝐛𝐢𝐥𝐢𝐭𝐲 𝐭𝐨 𝐀𝐮𝐭𝐨𝐧𝐨𝐦𝐲 [...]. Big bet. Some of it is still slideware, and I poke at that in the post. [...] That's the part that stuck with me. It isn't really Datadog's question. Anyone building infra runs into it sooner or later.
Full read:
#Observability #AI #AIOps #SRE #Datadog
Three Pillars history post:
"Metrics, logging, tracing." Anyone in observability can recite the three pillars like they're a law of nature. I went looking for where the phrase actually came from. The honest answer: 𝐍𝐨 𝐨𝐧𝐞 𝐝𝐞𝐬𝐢𝐠𝐧𝐞𝐝 𝐢𝐭.
[history: built separately, three problems...] Then commerce poured concrete over the crack. [...]
Here's the part that stuck with me.
[Bourgon drew the Venn diagram, then 18 months later wrote the 𝐨𝐩𝐩𝐨𝐬𝐢𝐭𝐞...] The industry grabbed the first idea and quietly ignored the second.
This is Part 1, on how it split apart. [caveat] [...] Worth remembering the next time someone tells you the current setup is just "how it's done."
Full read:
#Observability #SRE #Monitoring #DistributedTracing #DevOps
