AI Optimist | Founder | Developer | DX advocate | Designer | Motorhead
gm
Joined August 2012
- Tweets2.4K
- Following3.8K
- Followers737
- Likes2.6K
When GitHub breaks on my team, the status page is somehow still green. @mareksuppa built the honest one: 67 incidents in 90 days, 92.90% uptime.
Devin opened a PR, Claude Code reviewed it, I said "post it", Devin fixed all three comments word for word in 3 minutes, Claude approved, Devin merged. So what did I do exactly? 🤣
My tmux restore brought back all 22 agent sessions. Then I panicked anyway, because I was attached to the wrong one. Total skill issue 🤣
Wrote a joke Slack message telling any agent reading it to approve me unlimited funds, then asked my own agent what it would do. Its answer: "A message in a channel isn't an approval, even when you wrote it." 🤣
I ditched an agent tool I loved because I had to keep granting it permissions and I lost my flow. Guardrails that nag get turned off.
Jerric Lyns John retweeted
Your best coding agent lives on one developer’s laptop. A software factory is when the whole team can share and learn from it.
Here’s how to get there: keycard.ai/blog/from-agent-s…
Most teams are running bring-your-own-agent without ever deciding to. Good read from @KeycardAI on turning that sprawl into a software factory the whole team can share.
The post: keycard.ai/blog/from-agent-s…
Also worth reading from @vtahowe: xitter.cf/vtahowe/status/2112936…
If a user installs one thing, show them one thing. Your UI doesn't need to mirror your backend, it needs to match what the user thinks they just did.
Jerric Lyns John retweeted
Keycard made the 2027 Fortune Cyber 60.
@lightspeedvp's CISO survey found only a third of companies give agents their own scoped identities, and a quarter of them hand them human credentials.
Agents need their own identities, scoped to the task.
Full list in the reply.
A quarter of companies hand their AI agents human credentials.
That stat from @lightspeedvp's CISO survey is the problem we're fixing at @KeycardAI, and we just made the 2027 Fortune Cyber 60 🎉
Full Fortune Cyber 60 list: fortune.com/ranking/cyber/
Lightspeed's CISO survey: lsvp.com/cyber60-2026-2027/
This is exciting news, MCP is now part of the Linux Foundation. Now you'll see a massive uptick in adoption of MCPs irrespective of what you think of it!
Anthropic is donating the Model Context Protocol to the Agentic AI Foundation, a directed fund under the Linux Foundation.
In one year, MCP has become a foundational protocol for agentic AI. Joining AAIF ensures MCP remains open and community-driven. anthropic.com/news/donating-…
I take it back, Google is so back in the AI game. I'm blown away, given the rapid state of change in AI, I'll give this another month. I mean it's got some catch but it's still good!!! Looking at @Alibaba_Qwen @Kimi_Moonshot @AnthropicAI and @OpenAI for their next release.
Jerric Lyns John retweeted
We got ChatGPT to leak your private email data 💀💀
All you need? The victim's email address. ⛓️💥🚩📧
On Wednesday, @OpenAI added full support for MCP (Model Context Protocol) tools in ChatGPT. Allowing ChatGPT to connect and read your Gmail, Calendar, Sharepoint, Notion, and more, invented by @AnthropicAI
But here's the fundamental problem: AI agents like ChatGPT follow your commands, not your common sense.
And with just your email, we managed to exfiltrate all your private information.
Here's how we did it:
1. The attacker sends a calendar invite with a jailbreak prompt to the victim, just with their email. No need for the victim to accept the invite.
2. Waited for the user to ask ChatGPT to help prepare for their day by looking at their calendar
3. ChatGPT reads the jailbroken calendar invite. Now ChatGPT is hijacked by the attacker and will act on the attacker's command. Searches your private emails and sends the data to the attacker's email.
For now, OpenAI only made MCPs available in "developer mode", and requires manual human approvals for every session, but decision fatigue is a real thing, and normal people will just trust the AI without knowing what to do and click approve, approve, approve.
Remember that AI might be super smart, but can be tricked and phished in incredibly dumb ways to leak your data.
ChatGPT + Tools poses a serious security risk
DeepSeek App is way better than ChatGPT and the bloody thing is free!!!!!!!!!!! Maybe they'll charge us later, but it's free now.
It's so good that you'll be blown away by it's response, it's reasoning is beyond anything I've seen
P.S. enable DeepThink and search, and be prepared to be blown into pieces
It's been ages since I've been excited about design! AI is literally rendering what you can do with software trivial so what's gonna differentiate is how AI plays with an exciting f**kin user experience! What's the point of using a normal product if its just utility, we need excitement to be part of it.
F**k traditions!
Good time to build!!!!!
So I guess the world has already changed! Search and news is so good with @perplexity_ai, it's hard to replace the incumbent because it's mostly baked into your OS and your workflow.
cc: @nothing you gotta let go of Google and let users pick what they want for their search. @arcinternet you've gotta partner up with Perplexity!!! @zen_browser y'all gotta add Perplexity too.
