I've been using Claude Code heavily for the past few months while building my startup (12k+ line Python codebase), and it's genuinely changed the way I develop. It navigates large codebases with context that I haven't seen from any other tool: understanding file relationships, catching upstream/downstream effects of changes, and reasoning about architecture-level decisions, not just autocompleting lines.
But what really caught my attention is the business side. Anthropic is forecasting cash-flow positive by 2027, while OpenAI is projected to burn through $115B+ through 2029. Claude Code alone reportedly crossed $1B in annualized revenue. Their API revenue is estimated at double OpenAI's. And they're doing all of this with significantly lower burn.
The pattern is everywhere and I'd built it without thinking about it. Before the native App Store rating sheet, show your own little screen. Enjoying the app? Happy face goes through to the review sheet. Sad face goes to a feedback form that emails you instead. Every growth writeup calls it a best practice, and it does measurably raise your average rating. It raises it by making the negative reviews not happen. That's the entire mechanism. You're not collecting more sentiment, you're routing half of it into a private inbox where it can't reach anyone deciding whether to download. Once I'd said that out loud I couldn't keep it. What replaced it is dumber and I like it much more. Two positive moments an actual save, an actual "this helped" bank a signal. On the second one the native sheet fires directly. No pre-ask, no fork, no branch where I get to decide who's allowed to count. iOS makes the final call on whether the sheet even appears, which I have zero visibility into and have made peace with. Sixty-day cooldown, three prompts lifetime, and if the sheet isn't available on that device the signal stays banked instead of burning a prompt. Two things I got wrong on the way there. The first pass counted opening the app as a positive moment, which isn't a positive moment, it's a Tuesday it now only counts things the person actively chose to do. The second: the ask initially sat behind a paid surface, so the people most likely to be having a good time were the ones least able to be asked. Moving one of the two signals somewhere freely reachable mattered more than anything I did to the prompt itself. The part I'm still unsure about: I removed my own read on sentiment and replaced it with nothing. Unhappy people used to land in a form. Now they land in the App Store, or nowhere. Has anyone kept an always-available feedback door open without it quietly turning back into the sad-face branch?
but I recently came across an article describing how someone used Claude Code to access robot vacuum devices across 24 countries and potentially observe their environments.
I've been building in the AI-for-finance space, and I've noticed the real barrier isn't whether the models are good enough. They're getting scary good at reading documents, reconciling transactions, spotting patterns a human would miss. The barrier is trust and specifically, the gap between AI that reads your finances and AI that acts on them.
Here's the line I keep bumping into: people are surprisingly comfortable letting an agent categorize expenses or summarize their spending. But the moment it wants to do something file the taxes, move the money, recommend a five-figure decision the comfort evaporates. And I don't think that's irrational. The cost of a wrong email is an awkward apology; the cost of a wrong financial action is real money or an audit.
What's interesting is that the same person will draw that line in totally different places depending on stakes and reversibility. A $40 miscategorization? Whatever, fix it later. A mistimed refi or a missed deduction? Now trust matters enormously.
I continue building UI pages for my email marketing platform using AI, and I wondered if AI generated UI designs are good enough for production apps?
They seem to work great on mobile devices and desktop screens. They load fast. They behave normally. I m personally perfectly fine and happy to use them in production, but what is your opinion? Am I missing something? Should I be worried about using them in production? To build UI I use Lovable and Google Stitch. I go feature by feature, and for each feature I create a separate Git branch. This way I m more careful that one update does not break my entire website. P.S. The attached screenshot is a work in progress design that I created using Google Stitch for the Email Automation and Sequencing feature.
