The Hype Radar — score every launch in 90 seconds.

Every week, three or four posts surface on X, LinkedIn, Indie Hackers, Hacker News, all claiming the same thing. This will print money. This will replace your team. This will change your stack. Most are noise. A few will actually move what you ship. The cost of treating each one seriously — even briefly — is a tax on getting real work done. This is the framework I use to get to a verdict in 90 seconds. It's a working operator's version of Gartner's hype cycle — same instinct, ninety seconds instead of a year.

The problem

Two posts crossed my feed this week. One said ten open-source GitHub repositories were "quietly printing money" — fork them, white-label them, charge $200 a month. The other said six Claude Code plugins could "replace an entire dev team for free." Both had thousands of likes. Both came with a story arc — fork it this weekend, ship it next week — that has the shape of a deliverable but isn't one. The first looks like a business plan. The second looks like a tooling upgrade. Without a filter, both eat an afternoon.

I run three businesses — a Dubai consultancy called Articulate AI, a regional video pipeline for an insurance client, and a five-room sublease operation in my own villa. I have client work to deliver. The afternoon is not for sale. So before any deeper look, every hyped launch gets a 90-second pre-score on three axes.

The framework — P, Q, S

Three axes. Each scored 0 to 3. Total out of 9.

1. P — Personal productivity. Would I save hours per week by adopting this on Articulate, the client work, or operator infrastructure? Not "might" — would.
2. Q — Quality. Would output quality measurably rise — slop reduce, risk drop — on a real deliverable already in flight?
3. S — Sellable. Can it slot inside an Articulate workflow I bill for? Marketing Engine Pilot, Design Agent, security audit, content engine, data-sovereign stack.

Each score is one digit. Each digit is anchored to a real test. 3 means major — moves the needle. 2 means real — measurable benefit. 1 means marginal — possible edge case. 0 means no signal.

The reusable rule

Sum all three. Six or more is act. Install. Adopt. Build into Articulate. Four or five is conditional. Don't act yet. Define the trigger that would change the score (cost crossing a threshold, a client request, a second similar tool shipping) and re-score then. Three or fewer is skip. Don't waste cycles. Quarterly re-score, unless the category materially shifts.

One hidden rule. Items that clear 6 only on P or Q — install for yourself, do not try to sell. Only items scoring meaningfully on S become Articulate product surface. The number of times I've seen consultants try to monetise a productivity hack that solves only their own problem is what made the third axis non-negotiable.

Worked examples

Same week, same two viral posts. Different verdicts.

n8n — the open-source automation platform

Superpowers — the Claude Code plugin that forces plan-first behaviour

Claude-Mem — persistent memory across Claude sessions

gstack — Garry Tan's bundle of 23 role skills

The full table — all sixteen tools from both posts — is on the knowledge base. Four clear acts. Two conditional. Ten skips. The acts together form a coherent bundle I'm now packaging as Articulate's data-sovereign marketing stack: n8n + Listmonk + Plausible + the Anthropic Security Review plugin. None of the others combine like that.

Why this matters more in UAE

Eight UAE and GCC marketing agencies were surveyed alongside the launch posts. Traffic Digital, Tomsher, Digital Gravity, Born Creators Group MENA, Voliom Automate, Vista by Lara, AIST Riyadh, MAZ NEXA Riyadh. Two patterns emerged.

First, every competitor anchors on team size. 100-plus FTE. 200-plus FTE. Four sub-brands under a holding-co. Nobody inverts. Nobody sells "fewer humans + better AI = faster and cheaper." That's not a positioning gap I had to invent — it's untouched whitespace.

Second, "PDPL-clean" stack as a marketing headline is named by exactly one of the eight (Voliom Automate, narrowly, on the WhatsApp side) and never by anyone else. Federal Decree-Law 45 of 2021, DIFC and ADGM data residency, KSA's own PDPL — these are live buying concerns for any regulated mid-market UAE buyer. The data-sovereign bundle that fell out of this week's hype scan walks directly into that gap.

Put another way: the framework doesn't only filter noise. It surfaces where a UAE buyer's hidden anxieties match a tool I can actually install. That's how an evaluation discipline turns into a product decision.

How to use this yourself

Three rules of thumb.

One. The 90-second version is the version. Read the post, score P, score Q, score S, sum. If you're spending more than five minutes on a single item, the framework is failing — the score is the answer, not the analysis. The deeper read happens only on items that clear 6.

Two. Score S last. P and Q are easy to inflate. S is the discipline. If you can't name a client conversation in the next 60 days where this tool is the answer, S is 0 or 1. Most tools die at S.

Three. Log every score. Mine live in a single markdown file at ~/ClaudeWork/Claude/hype-radar.md with public excerpts on the knowledge base. Without the log, you re-score the same tools every three months because you forgot you'd already decided. With the log, the framework compounds.

The Sunday application of it is Stack Watch № 01 — first weekly sweep, scored, published. Sebastian runs the scan; the desk it runs from sits inside the launcher. Operator commentary I trust on the wider field: Simon Willison.

The next post in this series — The four-hour lead-response problem — covers UAE-specific data on why response speed dominates lead source, channel, and creative for SMB advertisers. Ships in June. Follow on LinkedIn for the drop.