Chief Technology Officer at Target Market Inc.
Greater Boston
Joined August 2023
- Tweets4K
- Following113
- Followers339
- Likes2.5K
Astro content collections now show up in Webflow Source's Content View, next to the @sanity_io integration.
Next on @webflow's list: custom agents and skills, and a first version of a faster CMS you can work in from Source. webflow.com/blog/source-note…
Pages with an author byline had 28% lower odds of being cited in a new study of 193,507 AI answers.
The author can't explain it and won't drop bylines over it. Neither will I.
I'd fix the opening instead. 20% of cited pages led with a direct answer. 5% of passed-over ones did.
Also in there: tables had 69% higher odds (88% in ChatGPT), a year in the title more than doubled them, schema was +13%.
And only 24-28% of cited pages got cited again a week later. Correlation only, and TJ Digital sells GEO services. tjrobertson.com/ai-citation-…
An agent hits a login page. It sends you a link, you sign in, click Hand back, and it keeps going.
@LangChain shipped that on Oct 29. The password never lands in the model's messages or traces.
That's how I'd want any agent working in a client's admin panel.
@LangChain open-sourced the design agent its team runs in @SlackHQ.
By their count it's made 3,000+ assets, and median cost per request fell from about $8 to $0.50 since mid-September.
I'd steal the brand skill first. Write your visual rules once and every request inherits them.
Company-reported, so grain of salt. Each Slack thread gets its own sandbox, and the repo says it takes over repetitive requests, not the designer. Code is Apache 2.0; hosting on Managed Deep Agents is paid. langchain.com/blog/design-ag…
Meta moved its new business terms from October 30 to November 20.
The crawl clause stayed in. Neither U.S. preview mentions robots.txt.
If you already set Meta crawler rules for Oct 30, leave them. You just got three more weeks for the client sites you didn't get to.
Meta says the extra time is to explain "what's changing and what's not," including how it uses crawlers. No regulator ordered the delay. WAN-IFRA called it "a welcome step" that "does not address our concerns." searchenginejournal.com/meta…
Webflow now shows where an AI agent is working on your canvas. Sparkle icon, highlighted element, and whose agent it is.
It doesn't lock anything. A teammate can still edit or publish mid-run.
I'd keep one rule anyway. Agents draft, a human publishes.
Marie Haynes' data via Search Engine Land. 145M+ sessions; conversion rates come from the 55 sites that track them. No industry mix or conversion definition given, so read the gap as directional. searchengineland.com/chatgpt…
TagetMarket's latest @webflow project is live: northeastdigestive.com/
Designed by Patrick Brickman, developed by yours truly.
I asked @grok to use historical design data from previous @newbalance models to predict what the navy and black colorways will look like. Anyone else want to contribute?
🤖 Made with AI
Researchers invented a concept, posted about it on open platforms, and waited.
Within 7 days, 8 of 10 AI search tools were citing it. @grok linked the planted posts in 30 of 32 answers.
I'd worry less about blog volume and more about where you show up.
Fudan preprint via SEJ, not peer reviewed. Seeds went on each tool's most-cited open platforms (@Medium, @YouTube). Topics were fictional, so it doesn't prove the same for real brands. @ChatGPT linked them in 7 of 20. searchenginejournal.com/ai-s…
Gemini mostly quotes the top of your page.
@resoneo went through 76,450 public Gemini conversations. A third of cited passages start in the first 10% of the page. 80% start in the first half.
If the answer sits under a 400-word intro, it's in the wrong spot.
Two more from the study:
51% of sourced answers cited at least one site outside the top 1M (or not in CrUX at all), and median searches per exchange fell from 4 to 2 this year.
The authors label their "picks the page, then the passage" model as inference.
think.resoneo.com/gemini-con…
Claude Code Review is out of preview for Team and Enterprise.
Most reviews now finish in 6-10 minutes, down from about 20 in preview.
Before you turn it on, know the default. New repos get reviewed after every push, the most reviews you can run. I'd start on Manual.
Manual means nothing runs until someone comments @claude review on the PR (or @claude review always to follow every push). Drafts and forks get skipped either way. Setup is in Admin settings > Claude Code > Code review. code.claude.com/docs/en/code…
