Matt C. A. Smith
@mattcasmith.net
💻🔑 By day, I stop cyber attacks. 🗒️✍️ In between, I write sporadic field notes from the intersection of technology, design, and life. Personal posts.
mattcasmith.net
mattcasmith.net
AI browsers like ChatGPT Atlas will shape the future of the web, but routing access through an intermediary should be concerning to everyone. [7/7]
ChatGPT Atlas positions AI as the internet's gatekeeper | MattCASmith
We’re shoehorning AI into everything else, so why not your web browser? That was presumably the thinking when OpenAI launched ChatGPT Atlas, an AI-augmented browser that it says will change the way we...
mattcasmith.net
November 5, 2025 at 2:05 PM
AI browsers like ChatGPT Atlas will shape the future of the web, but routing access through an intermediary should be concerning to everyone. [7/7]
Social media platforms already sanitised the Wild West internet of the 1990s and 2000s. AI browsers take the the idea further, towards a smaller, safer web that runs along corporate guardrails. [6/7]
November 5, 2025 at 2:05 PM
Social media platforms already sanitised the Wild West internet of the 1990s and 2000s. AI browsers take the the idea further, towards a smaller, safer web that runs along corporate guardrails. [6/7]
Worse is that LLMs already refuse to discuss anything moderately controversial. If AI sits between you and the web, that filter becomes a barrier and parts of the internet effectively cease to exist for you. [5/7]
November 5, 2025 at 2:05 PM
Worse is that LLMs already refuse to discuss anything moderately controversial. If AI sits between you and the web, that filter becomes a barrier and parts of the internet effectively cease to exist for you. [5/7]
Even my LinkedIn feed is starting to unearth examples of prompt injection, where malicious websites with hidden content could trick AI browsers into performing actions or handing over private information. [4/7]
November 5, 2025 at 2:05 PM
Even my LinkedIn feed is starting to unearth examples of prompt injection, where malicious websites with hidden content could trick AI browsers into performing actions or handing over private information. [4/7]
Also, think about how unreliable this generation of AI is in general. Do you really want to give tech that can't count letters in a word access to your logged-in bank account - even under your supervision? [3/7]
November 5, 2025 at 2:05 PM
Also, think about how unreliable this generation of AI is in general. Do you really want to give tech that can't count letters in a word access to your logged-in bank account - even under your supervision? [3/7]
First, AI just isn't that good at anything requiring context yet. The demo shows ChatGPT buying beach supplies, but IRL it won't know what Jimmy's favourite snack is, or what toys the kids already have. [2/7]
November 5, 2025 at 2:05 PM
First, AI just isn't that good at anything requiring context yet. The demo shows ChatGPT buying beach supplies, but IRL it won't know what Jimmy's favourite snack is, or what toys the kids already have. [2/7]
Beyond the financials, there's a whole lot of interesting stuff about the relationship between Anthropic, Cursor, and Claude Code in his latest podcast and newsletter. www.wheresyoured.at/costs/
This Is How Much Anthropic and Cursor Spend On Amazon Web Services
So, I originally planned for this to be on my premium newsletter, but decided it was better to publish on my free one so that you could all enjoy it. If you liked it, please consider subscribing to su...
www.wheresyoured.at
October 21, 2025 at 7:57 AM
Beyond the financials, there's a whole lot of interesting stuff about the relationship between Anthropic, Cursor, and Claude Code in his latest podcast and newsletter. www.wheresyoured.at/costs/