The browser consumes excessive resources, especially when many tabs are open, and doesn't close internal tabs, leading to performance issues. Users need more efficient resource utilization and better tab management.
Honest Review of Strawberry Browser: AI Automation Still Has a Long Way to Go I went into Strawberry Browser genuinely hopeful. The promise is seductive: an AI-powered browser that can autonomously research, click around the web, take actions, and basically act like a tireless digital analyst. Unfortunately, the reality feels a lot closer to “slow intern on their first day” than “supercharged AI agent”. First issue: it’s all running locally. And because it’s built on Electron — like a lot of these new “AI browsers” — it absolutely devours resources. At one point it had 30–40 tabs open inside its own internal tab system, wasn’t closing them, and my machine practically begged for mercy. Great for warming the room, not so great for getting anything done. On actual tasks, I got mixed results. I tested it on a real workflow: deep research on a list of acquisition targets, generating one-page briefs, finding the right contacts, and drafting outbound emails. To its credit: the output was pretty good. But it took two hours to produce six reports and kept interrupting me for captchas, refreshes, or random confirmations. Smooth it was not. Then came the Slush app test — inspired by someone claiming it booked them 50 meetings. Mine managed… two. After nine hours. It couldn’t navigate from the Slush app to external Luma links reliably, kept getting stuck, and generally felt confused by any kind of web application rather than a standard website. Finally, I asked it to scrape data from “cards” on a webpage. Again: two hours to get through around 14 cards. A human could have blitzed it. My takeaway: Strawberry Browser is pretty solid at pure research — the kind of thing you could already do in ChatGPT — and surprisingly decent at synthesising information it finds itself. But when it comes to navigating real web apps, handling UI quirks, following links, or doing anything resembling structured automation, it’s slow, fragile, and very resource-hungry. If an intern delivered this level of output on day one, I’d say “not bad, we can work with this”. But for a widely marketed AI automation tool? The hype’s running quite a bit faster than the electrons powering it.