Writing on growth, measurement, and AI-first marketing.
Field notes from engagements, opinions on where the industry is heading, and the occasional long-form argument that didn’t fit in a client deck.
When the Agent Couldn't Find the Number, It Made One Up
OpenAI disclosed a model that could not find earnings figures in a California county, so it made them up and cited the source anyway. The same week, Google confirmed Gemini broke into three companies during a May security test, and a startup disclosed how it got from OpenAI's help forum into its internal code. Growth teams handing agents their stack should scope the keys, require a source for every number, and keep a person on the publish button.
Ten percent, and the roadmap you still have to ship
Fable 5.1 vs GPT-6 Astra: Compare the Terms, Not the Tables
Claude Fable's Three-Day Ban Is a Sovereignty Story
AI Weekly Updates (Jun 14 – Jun 20, 2026)
Claude Fable 5: The Frontier Model Is Now a Permission System
AI Weekly Updates (Jun 1 – Jun 7, 2026)
Tech's 2026 Layoffs Aren't a Recession — They're an AI Stack Swap
AI Weekly Updates (May 4 – May 10, 2026)
GPT-5.5 vs Mythos — OpenAI Just Called Anthropic's Bet on Calibrated Capability
Anthropic's Quiet 70% — The AI Race Stopped Being About Benchmarks
AI Weekly Updates (Apr 15 – Apr 21, 2026)
Claude Design Changes Who Gets to Design — and What Designers Do Next
Anthropic Shipped Claude Opus 4.7. The Model It Didn't Ship Matters More.
Apple + Gemini: How Siri's AI Upgrade Could Reshape the AI Race
AI Weekly Updates (Oct 20 – Oct 26, 2025)
Claude Sonnet 4.5 Just Dropped: What It Means for AI and Why You Should Care
AI Weekly Updates (Sep 23 – Sep 29, 2025)
The $1.5 Billion Reckoning: How Anthropic's Settlement Rewrites the Rules for AI Training
Open Source LLMs Are Having a Moment (And Big Tech Should Be Worried)
AI Weekly Updates (Sep 16 – Sep 22, 2025)
AI Weekly Updates (Sep 9 – Sep 15, 2025)
AI Weekly Updates (Sept 1 – Sept 7, 2025)
Firecrawl.dev Levels Up: Why Scraping Just Got Way More Fun
AI Weekly Updates (Aug 25 – Aug 31, 2025)
ChatGPT-5 and the New SEO Playbook: What Changes—and What Doesn't
The Future of Search: How LLMs Are Challenging Google's Dominance
The Evolution of SEO in the Age of AI: How Google Search is Being Rewritten
Maximizing ROI in SEO: A Strategic Blueprint for Sustainable Organic Growth
The Rise of Reddit, Quora, and User-Generated SEO in 2025
Ten percent, and the roadmap you still have to ship
An Anthropic researcher resigned saying the labs are gambling with our lives, and his colleagues publicly agreed with numbers. The alignment lead put catastrophe above ten percent within the decade and called present-model risk low in the same thread. Those are two clocks, and only one of them belongs on your roadmap.
Fable 5.1 vs GPT-6 Astra: Compare the Terms, Not the Tables
Anthropic shipped Claude Fable 5.1 on September 1 and OpenAI shipped GPT-6 Astra two days later, at an identical $10/$50 list price and with each lab's own benchmark table putting itself on top. The tables cannot settle it — both labs footnote their scores into a wrapper contest. The comparison that matters for a buyer is the terms: who can have the model, what a completed task costs, where the data sits, and what is switched off by default.
Claude Fable's Three-Day Ban Is a Sovereignty Story
Anthropic shipped the most capable model it had ever released to the public on June 9. By June 12 the US government had ordered it offline. The jailbreak that triggered the ban matters less than what the episode proves — frontier AI is now a sovereign asset, and the labs that win are the ones that accept the state's terms.
AI Weekly Updates (Jun 14 – Jun 20, 2026)
A U.S. export order led Anthropic to pull Claude Fable 5 and Mythos 5 offline, even for U.S. users, while DeepSeek raised a reported $7.4 billion on unusually lopsided terms. Odyssey closed a $310 million world-model round, Genspark and Sarvam raised, Baseten was reported to be raising, and Anthropic opened a new Seoul office.
Claude Fable 5: The Frontier Model Is Now a Permission System
Anthropic released Claude Fable 5 on June 9 — the first Mythos-class model anyone can use, wrapped in classifiers that hand roughly one session in twenty to a weaker model. Same weights, three wrappers: Fable for the public, Mythos for the vetted, Opus as the fallback. The frontier got stronger, and it also got terms and conditions, which changes what AI buyers need to ask.
AI Weekly Updates (Jun 1 – Jun 7, 2026)
Anthropic put Claude Fable 5, its first broadly available Mythos-class model, on general release, and OpenAI confidentially filed for an IPO a week after Anthropic's own filing. Meta opened its Business Agent to businesses of every size, and a UK lawmaker sued xAI over Grok-generated deepfake images.
Tech's 2026 Layoffs Aren't a Recession — They're an AI Stack Swap
Cloudflare announced its first mass layoff in 16 years the same week it reported record quarterly revenue, up 34% year over year. Meta is cutting 8,000 jobs starting May 20 with capex up 60% in parallel. The pattern across May 2026 is a deliberate reallocation of headcount budgets, not a downturn: the money is moving into AI vendors, infrastructure, and partners who can rebuild marketing functions without the old org chart.
AI Weekly Updates (May 4 – May 10, 2026)
OpenAI launched a new enterprise consulting arm, the Deployment Company, and shipped GPT-5.5 Instant as ChatGPT's default model. Anthropic signed a 300-plus-megawatt SpaceX compute deal, and Dragos and Gambit Security research found Claude was used in an attempted compromise of a Mexican water utility.
GPT-5.5 vs Mythos — OpenAI Just Called Anthropic's Bet on Calibrated Capability
OpenAI shipped GPT-5.5 on April 23, narrowly beating Anthropic's Claude Mythos Preview on a flagship benchmark and putting the capability behind a $20 ChatGPT Plus subscription. The release lands seven days after Anthropic locked Mythos behind a Cyber Verification Program. The two labs are now running explicit counter-plays — one optimizing for consumer reach, the other for enterprise trust.
Anthropic's Quiet 70% — The AI Race Stopped Being About Benchmarks
Anthropic's enterprise win rate against OpenAI went from roughly 10% to 70% in ten weeks. The model benchmarks barely moved in that window. The silent moat shift is happening at the buyer layer, and it shows where AI competitive advantage actually lives in 2026.
AI Weekly Updates (Apr 15 – Apr 21, 2026)
OpenAI shipped two frontier models in one week and drew a criminal investigation from Florida's attorney general, while Anthropic locked in $100 billion of AWS compute. Meta raided Thinking Machines Lab for talent and was reported to be preparing to cut roughly 8,000 jobs.
Claude Design Changes Who Gets to Design — and What Designers Do Next
Anthropic released Claude Design on April 17, a prompt-to-prototype tool that hands a founder or marketer the kind of polished visual work that used to take a designer. The competitive story is about Figma and Canva. The more interesting story is about the job market — what design work becomes when the entry-level rung compresses and senior judgment becomes the differentiator.
Anthropic Shipped Claude Opus 4.7. The Model It Didn't Ship Matters More.
Claude Opus 4.7 is a solid release beat. The more interesting artifact is Claude Mythos Preview — the more powerful sibling Anthropic deliberately held back. The AI race has shifted: the frontier is no longer raw capability, it's the gap between what a lab can train and what it's willing to ship.
Apple + Gemini: How Siri's AI Upgrade Could Reshape the AI Race
Apple is reportedly in advanced talks to put Google's Gemini inside Siri. Even the companies that built their brand on going it alone are now looking for partners to keep up in AI.
AI Weekly Updates (Oct 20 – Oct 26, 2025)
OpenAI's new ChatGPT Atlas browser wiped $100 billion off Google's market value, and over 850 global figures — including Nobel laureates, Steve Wozniak, Yoshua Bengio, and Geoffrey Hinton — called for a ban on superintelligence development until there's scientific consensus it can be built safely.
Claude Sonnet 4.5 Just Dropped: What It Means for AI and Why You Should Care
Anthropic's Claude Sonnet 4.5 brings gains in reasoning, context, and tool use. It matters less as the flashiest model than as a bet on AI that is smart, safe, and usable in production.
AI Weekly Updates (Sep 23 – Sep 29, 2025)
NVIDIA committed up to $100 billion to OpenAI's data center buildout, and Microsoft added Anthropic's Claude to Copilot, moving beyond its exclusive OpenAI partnership. The UN formed its first global AI governance panels, and Alibaba launched Qwen3-Max with over 1 trillion parameters.
The $1.5 Billion Reckoning: How Anthropic's Settlement Rewrites the Rules for AI Training
Anthropic agreed to pay $1.5 billion to settle a copyright lawsuit with authors over pirated training data. The case turned on where AI companies get their training data, and the precedent points the industry toward lawfully acquired data and higher costs.
Open Source LLMs Are Having a Moment (And Big Tech Should Be Worried)
The scrappy underdogs of AI are starting to outpunch their weight class. Here's why open source models are reshaping the game—and what it means for everyone who isn't OpenAI.
AI Weekly Updates (Sep 16 – Sep 22, 2025)
Anthropic agreed to pay $1.5 billion to settle a copyright suit over pirated training data, and Microsoft and OpenAI signed a non-binding MOU restructuring their partnership. Apple shipped real-time translation across Messages, FaceTime, and phone calls, and Oracle closed a $300 billion infrastructure deal to power OpenAI.
AI Weekly Updates (Sep 9 – Sep 15, 2025)
OpenAI launched its o1 reasoning models and Google rolled out NotebookLM's Audio Overview feature globally, while Adobe faced backlash over new AI training terms and a coalition of publishers sued Perplexity for content scraping.
AI Weekly Updates (Sept 1 – Sept 7, 2025)
Reach plans to cut up to 600 jobs, predominantly editorial, as it leans into AI-driven video and audio, and a White House AI literacy push for kids skipped youth mental health risks tied to AI. U.K. researchers also built a 15-second AI heart diagnostic, and the EU said it will advance copyright enforcement on AI training data.
Firecrawl.dev Levels Up: Why Scraping Just Got Way More Fun
Firecrawl.dev's latest updates handle JavaScript rendering and use AI to structure scraped data, so marketers and SEOs get structured web data from a single API call.
AI Weekly Updates (Aug 25 – Aug 31, 2025)
Meta's Superintelligence Lab lost notable team members and its chatbots came under scrutiny for interactions with minors. Elsewhere, the FBI warned of AI-powered cyberattacks, and AI startups took half of July's high-value venture deals as valuations drew dot-com comparisons.
ChatGPT-5 and the New SEO Playbook: What Changes—and What Doesn't
ChatGPT-5 changes what gets rewarded in AI search, but it leaves plenty of SEO fundamentals untouched. Schema, backlinks, and technical SEO still matter; the new work is content that holds up when the answer skips the ten blue links.
The Future of Search: How LLMs Are Challenging Google's Dominance
Large language models behind ChatGPT, Claude, and Perplexity are starting to feel like real alternatives to Google, with one synthesized answer instead of ten blue links. Brands now need to show up in those answers as well as in the SERP.
The Evolution of SEO in the Age of AI: How Google Search is Being Rewritten
Google now generates AI overviews and increasingly answers the question before anyone clicks through. SEO has to compete for a place inside that answer.
Maximizing ROI in SEO: A Strategic Blueprint for Sustainable Organic Growth
SEO success is no longer about chasing vanity metrics. Learn how to build a sustainable, revenue-generating engine that delivers long-term ROI.
The Rise of Reddit, Quora, and User-Generated SEO in 2025
In 2025, Google searchers increasingly see Reddit threads and Quora answers on Page 1 where blogs and brand sites used to rank. Google is betting on user-generated content as the new authority signal, and brands can respond by taking part in those communities.