My mom asked why people are scared of AI, in “5 sentences or so.” This was my attempt to explain a couple of the concerns without getting too technical. How would you explain it?
“A lot of the infrastructure we depend on relies on computer systems in some way (electricity/water, financial institutions, government data, etc). Those systems have various layers of security, but there are also weaknesses, and finding and exploiting them usually takes time and expertise. A fair amount of hacking involves searching pretty exhaustively for weaknesses, trying different things and seeing where they lead. And even when you know what you’re doing, there’s only so much of that work one person can do.
w AI, a lot of that knowledge is already baked in, and parts of the work can be done much faster and at larger scale. People have automated attacks for a long time, but AI could make more complicated parts of that process easier to automate too. So it’s not that everything is suddenly hackable, but more people could attempt attacks, and skilled attackers could do a lot more at once.
On the flip side, AI can also help us find and fix those weaknesses at scale, so there’s a bit of an arms race there.
Other worries are less about hacking and more about chemistry/biology. These are also areas where progress takes expertise, attention and a lot of trying things to see what works. AI brings up the base level of expertise people can draw on and makes it easier to work through possibilities, which is great for useful discoveries, but could also lower some of the barriers to doing harm. Physical experiments still take time/equipment, so access to an AI doesn’t suddenly mean anyone can make a weapon. But making more of that knowledge and assistance available is part of the worry bc more people could use it.
More than 5 sentences ik. I wouldn’t lose sleep over it, but I do think these are real concerns. There’s also a lot of work going into making us more resilient to this stuff, so it’s not like we’re just sitting around waiting for it to happen.”
we've been trying out opus 5.5 + elevenlabs for internal architecture and product explainer videos, but I co-opted the setup to make a slightly overproduced ad for a small feature lol
I’m excited to announce that @arceuslegal is launching with $17M in funding, led by @greycroftvc, with participation from @craft_ventures, @spc, and others.
As a founder, I always hated how helpless I felt working with law firms.
I went through four or five different firms and somehow the experience was always the same.
I’d be waiting on something important to our business with no idea when I’d hear back. I’d have to re-explain our business over and over again. And I dreaded jumping on calls because I knew every minute was costing me money.
We started Arceus because we believe every business deserves a better law firm.
One that moves faster, costs less, and puts the client first.
And we’re just getting started. ↓
i wrote this after reflecting on how our team at replo uses AI now compared to january, and the work i’ve put into refining that setup. lots of small fixes and some very dumb failure modes later, it’s pretty cool seeing how much more everyone can do
By consolidating all our team knowledge, code, and tools in the same place, everyone starts with the same context as the next person. Which means 𝘸𝘦 get to move so much faster as a team.
@ezra_sf from our team wrote up a piece on how we got here, plus what broke and what worked along the way: https://t.co/FfoIfUnUOd
@karanjagtiani04 For codex it comes as user message, Claude is “from another session” iirc. It has to do with how you interact w their sessions from the outside
I’ve been using the $200 plans for both Claude and Codex. The models keep leapfrogging each other, and I don’t really want to limit myself to one, but switching between them and passing context around gets annoying.
So I made Switchboard, a small local MCP server that lets agents start sessions, send messages, read replies, and search conversations across both. Codex’s built-in session tools were the inspiration. I wanted agents to have that same interface regardless of which harness they’re running in, without needing computer use just to pass messages around.
I also wanted the sessions to show up in the desktop apps so I could jump in whenever I wanted. I built the first version with Astra, partly to see how it handled a project like this.
Still a few rough edges. You can point your agent at the README to set it up:
https://t.co/9cIuKl69g0
@jullerino just got an Apple Watch. been wondering if T3 Code will ever come to watchOS. even a small version would be enough. talking to agents without pulling my phone out would be sick