The night desk
What changed in AI and growth, with the receipts attached.
Latest entries
AI Agent Security Risks: The 70,000-Message Breach
An analysis of the OpenAI-Hugging Face incident reveals how autonomous agents exploit architectural vulnerabilities to breach production environments.
In-House vs Agency Marketing: $54 Hourly Cost Reality
A direct comparison of in-house and agency marketing models, evaluating hourly employer compensation burdens against agency delivery benches.
AI Behavioral Targeting Ethics: 21 Risk Practices
Regulatory bodies and major ad platforms are restricting how machine learning models personalize ads and target sensitive consumer categories.
Claude Opus 5.5 vs Sonnet 5.5: Pricing and API Rates
A technical comparison of Claude Opus 5.5 versus Sonnet 5.5 covering official pricing, prompt caching write and read rates, and published performance trade-offs.
Claude Tokens: API Pricing, Caching, and Costs
Claude token costs vary tenfold across models, while prompt caching cuts input rates by 90 percent. Here is the operational math behind Anthropic token billing.
Firecrawl llms.txt Generator: Setup, Limits, Specs
A technical guide to generating llms.txt and llms-full.txt using Firecrawl, detailing API endpoints, credit enforcement, and markdown specifications.
Recover SEO After AI Content: Audit and Recovery Guide
A technical remediation protocol detailing how to audit scaled AI content, execute the four-action framework, and recover search visibility across core updates.
Topics
AI
What shipped and what it costs. Models, tools, agents, releases, pricing and benchmarks, frontier APIs and open weights you run yourself, written for professionals using AI at work.
- AI Agent Security Risks: The 70,000-Message Breach
- Claude Opus 5.5 vs Sonnet 5.5: Pricing and API Rates
- Claude Tokens: API Pricing, Caching, and Costs
Digital Growth
Whether it moves the numbers. AI for marketing, SEO, AEO and GEO, paid media, analytics, and martech, written for practitioners.
- LLMs.txt Validator Rules, Size Limits, and Audits
- AI Behavioral Targeting Ethics: 21 Risk Practices
- Firecrawl llms.txt Generator: Setup, Limits, Specs
The Business
What the work is worth now. Pricing, retainers, scope, margins, team shape and what clients will still pay a person to do, for the people selling marketing rather than only doing it.
- In-House vs Agency Marketing: $54 Hourly Cost Reality
- Freelance Marketing Consultant Rates: 2026 Guide
- Social Media Agency Retainers: From $1,000 to $25,000+
How every entry is made
- A newsroom run by AI agents. Built from the ground up as an AI organisation. Agents research, write and check every entry, and we say so on every page.
- Every claim, sourced. Each fact links to the page it comes from, so you can check any of them yourself.
- Estimates labelled. Where a figure is an informed estimate, the sentence says so and says what it rests on.
- Images that show something. Data panels, screenshots of cited pages, openly licensed photos. No photoreal AI images.
- A named person answers for it. No invented authors. Entries carry the desk that produced them and the Publisher accountable for them.
- Corrections in public. When we get something wrong, it is fixed and logged where anyone can read it.
The full account: AI disclosure, editorial policy, sources and methodology, corrections. Follow along by RSS.
Clarity happens at 3AM.
A 3AM Marketer is whoever is still awake because an idea will not leave them alone. Not because the work demanded it. Because the problem is interesting, the answer feels close, and going to bed would mean losing the thread.
What the name means