voice is cheap to fake now and caller id can be spoofed, so a call that sounds like your mom or your boss or your bank proves nothing on its own
if a call asks for money or a code, hang up and call them back on a number you already have
for family i wanted something quicker, so i built vouchd. your family shares a code that keeps changing, same idea as the 2fa app on your phone. if the voice cant read you the current code, its not them
agents are going to start making legit calls for people too, so at some point calls need some kind of signed proof of who sent them, or people just stop picking up
altman said at stanford he no longer thinks were going to be able to just patch all the bugs. i think hes right
if finding a bug gets cheap, you have to assume some piece of your system is already broken and decide what that piece is allowed to do
for anything that moves money that means each service gets its own keys and its own limits, and nothing posts to the ledger unless it reconciles. a bug in one place should cost you one bad afternoon
it also argues for less code in general. each dependency you skip is one less thing a model can go read for holes
ai made custom software cheap to build. someone still has to keep it running
so id bet on custom software as a service. a shop builds something just for your company, then gets paid monthly to patch it, fix the integration when a vendor changes an api, and still know how it works after the person who prompted it moved teams
id happily pay for software that fits the business with someone on the hook for it
mining might be the cleanest place to let ai agents run real hardware. one number tells you if it made things better or worse
everyone wants to know where bitcoin will be next year. almost nobody asks where it'll be in 20, and thats the only answer that ever matters
read the order. it doesnt touch rates, it goes after the costs stacked on top of them.
the interesting part is pmi. right now it comes off based on what you originally paid for the house. this would have fannie and freddie check what its worth today with automated valuations and drop it once you have the equity. thats a data problem more than a policy one, and probably the most useful thing in here.
the catch is it only applies to loans made after the pilot launches, and the title savings only show up on refis. with rates above 7 not many people are refinancing. helps whoever buys next more than whoever is paying today
meanwhile half the industry is handing agents a browser, a terminal and an api key and calling it a product launch.
im not saying stop. im saying build like the lab does. assume every permission you hand an agent will get tested, every sandbox probed, every misconfig found. not because it wants to cause harm, because it never gets tired of trying.
most security was built for attackers who eventually go home. this one doesnt
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
coding agents now get a welcome back email about their dependencies
stalemap is an open-source tool that tells coding agents which dependencies changed after their training cutoff.
→ It compares your pinned versions with those from before the model's cutoff, for JS and TS projects
→ It writes notes into AGENTS.md or CLAUDE.md and adds an MCP search tool
→ It checks edits with the project's type-check and flags stale API use
→ In the author's own test, 89% of sessions passed with it versus 68% without
Repo: github.com/Umair/stalemap
the private key was 4. just 4
So basically, today PHP got two critical bugs that let anyone run code on your server without even logging in, which is bad, and then the details leaked before the patch was even out, which is worse, and it works on WordPress and MediaWiki and Magento and Joomla and Laravel, and Censys says like 1.5 million servers are running a version with no fix, so the NSA was like "take your system offline," and the FBI was like "take your system offline," and CISA and the UK and CERT-EU were all like "take your system offline," and then the Prime Minister of Japan, the actual Prime Minister, posted at 5:55 in the morning telling 48.8 million people to take their system offline, so everyone took their system offline, like the Dutch cybersecurity agency took its own website offline, and the Dutch parliament went offline, and the Port of Rotterdam went offline, and Finland and France and Denmark started pulling government sites, and Wikipedia went down, and Archive.org went down, and Asana and GOG and Something Awful went down, and the University of Vienna's webmail went down for some reason, and that's just PHP. Because GitLab also dropped a CVSS 10.0, which is the highest score they give out, for a bug that's been there since 2014, so that's every self-hosted GitLab for twelve years, and then LiteLLM 1.107.0 got hijacked and shipped a credential stealer and a remote access trojan in every single wheel and the Docker images, including latest, and you didn't even have to import it, just installing it was enough, and it's the second time this year, and then Railway paused every deployment on the entire platform because someone got into the environment variables, so now everyone has to rotate every secret they've ever stored there, and TiDB Cloud said someone was walking around inside with stolen credentials for basically a month and an extortion group says they took 10.6 terabytes, and Zeabur got breached for the second time in two months and is just shutting down, like the whole company, and Node.js patched remote code execution and a TLS bypass, and CISA says Plex is being actively exploited, and someone took over 300 Tor exit relays to steal SSH passwords, and SafePay ransomware has hit over 600 critical infrastructure orgs, and India's biggest bank filed a data loss incident, and Japan has been in an official state of emergency in cyberspace since October 9, which honestly explains the 5:55 AM post. Oh, and eMudhra, a certificate authority that's issued over 60 million certificates, whose entire job is proving websites are who they say they are, was supposed to be using private keys that are random numbers over a hundred digits long, and one of them was 4. Just 4. And the others were like 66 and 237. Anyway, it's Wednesday.