# Marcus Tran, Lead Engineer at Formstack-sized B2B SaaS (~60 engineers) — read of Deep Work, June 12 2026 > "8 years writing TypeScript, currently tech lead on a payments squad, VS Code all day, GitHub Copilot since 2023, Claude since last fall. 3-year-old daughter. I code from 9pm to midnight four nights a week and I am acutely aware of every minute I waste." --- ## How I got here Someone in the #tools channel at work dropped a link with the message "anyone tried this?" No context, no endorsement. I clicked it because I was already having the exact bad morning described in the first paragraph: I had Alt-Tabbed between my editor, Claude, a Confluence doc, and Slack 11 times in one hour before I'd written a single meaningful line. The timing was unfortunate for my skepticism. ## What I clicked first The stat block stopped me. "40m Lost to context switching daily" and "8x More interruptions with AI tools." I actually sat up a little. Not because I believed them immediately, but because the second one matches something I've been quietly noticing since I added Claude to my workflow. The AI tools make individual tasks faster but they pull me out of the editor constantly. That tension is real. So the thesis is real. I kept reading. ## Where I paused "The tool detects when you have real momentum and protects it." That's the Smart Pauses feature. I paused here for a while. How? Keystroke velocity? Git diff size? Cursor idle time? I genuinely want to know the mechanism because the claim is doing a lot of work. If it's just "no notifications for 25 minutes" then it's a Pomodoro timer with extra branding and the word "detect" is misleading. If it's actually doing something interesting with editor telemetry, that's the whole product and there's not a single word about how it works. ## What I distrusted Three things. First: "Copilot + Deep Work = 40% faster feature delivery (early metrics from beta)." That sentence is shaped like a data point but it's empty. 40% faster than what baseline? Measured how? Over what period? "Early metrics from beta" could mean two people filled out a survey. I've seen this kind of stat used on landing pages to manufacture credibility and it does the opposite for me. Second: "Join 500+ developers in the Deep Work beta." Fine, but then the pricing block says "Free for the first 50 teams through 2026." Those two numbers are inconsistent enough to make me wonder if both are made up. Third, and this one genuinely confused me: there's a whole section at the bottom that says "How honest is this idea, really?" with scores like "financial upside: 1/10" and "Year-1 take-home: -$12,000." That is not product copy. That is a business-idea marketplace listing showing me the founder's Fermi estimate of how little money this will make. Why is that on the product homepage? I read it twice thinking I was missing something. It makes the page feel like I accidentally opened the wrong tab. I'm a potential user, not a potential buyer of the business idea. This section broke the spell completely. ## What would convince me Show me a screen recording of the "Integrated AI Panel" actually working inside VS Code. Not a mockup. A real session where someone is coding, hits a question, pulls up Claude in a side panel without leaving the editor, gets an answer, and keeps going. That one thing would answer half my questions. For the 40% claim: one specific team, company size, stack, what they measured, before and after. A Loom from their lead engineer would do it. I don't need a case study PDF. I need a human saying "here's what changed for us." ## What I'd ask in an email reply 1. The Smart Pauses description says the tool "detects when you have real momentum." What signals is it actually reading? Keystroke cadence? Time since last file save? I want to know if this is real ML or just a marketing description of a timer. 2. The Integrated AI Panel connects to Claude, Copilot, and ChatGPT. Does this require me to authenticate with each service separately and does it use my existing API keys? Or is it a proxy? What's the data path for my prompts? 3. The bottom of the page has a section showing this product's own "adoptability score" and a negative Year-1 Fermi estimate. Is this page also a listing in some kind of idea marketplace? I'm confused about whether you're selling me the tool or selling me the idea to build the tool. ## Verdict: on-the-fence The problem statement is real and I felt it this morning. But the page confused me badly with the idea-marketplace section at the bottom, the inconsistent beta numbers, and the hollow "40%" claim. If someone emailed me with a screen recording of the VS Code panel working, I'd reply within an hour. --- *Memo by skeptic persona, generated 2026-06-12. Studio breaks own self-grading loop.*