Robert Hafner

The AI RCEssion is driven by tech debt, but that debt is not evenly distributed. AI isn't the adversary coming to destroy us, it's the auditor pointing out our decades of neglect and lack of investment in secure systems.

This is why I bought a GPU machine. It took less than four months running open models to save back that money (yes, including electricity) instead of paying the vendors

Where's the list of the demos who didn't sign it? We need to know who to primary next round.

PHP remote code execution, intermediate certificates with private keys that are a single byte, what a week to be alive

Remember folks, when the big companies steal all of your data and use it to train models it's innovation. When Chinese companies do the same thing to the big companies it's an "attack".

The old model was cheaper batch processing. You could simply move your LLM calls to an async model, wait 12-24 hours, and save 50%. This allowed the providers to load shift. That doesn't work in the agentic world though, where a session can have dozens or hundreds of calls that have to go in order.

If you haven’t asked abliterated GLM 5.3 to hack the Internet’s core routers, then you’re a happier person than I am right now.

"I tried running it past comms but they wouldn't let me post stupid shit so now i just ignore them" is a hell of a take

Two bad migraines this week, just need to shout that out to the void

New: We obtained witness videos and surveillance footage showing ICE parking their cars, arresting an observer, and transferring detainees in parking lots across 4 of Chicago's police districts. Local officials have tried banning ICE from city property. But incidents in the lots keep stacking up:

Beyond the Vibes: A Rigorous Guide to AI Coding Assistants and Agents has been updated! A lot has changed since I first wrote this, but the need for quality learning material has not. This is my attempt to give developers some grounding as they jump into this space. blog.tedivm.com/guides/2026/...

When you see this ask what the owners of CBS are getting out of it. They aren't just throwing away viewers (and advertising) without expecting to profit in some other way.

My AI predictions for the next two years: 1. Carbon emissions will increase 2. Wealth inequality will increase 3. Movies, television, and other media are going to get a lot shittier

The AI slowdown is just the big companies attempting to build a moat around their failing businesses

This is horrifying. We've had heat waves, storms, power outages, downed trees, and wildfire smoke from the north. If we're the most resilient city on earth then it's hard not to think that we're absolutely fucked as a species. blockclubchicago.org/2026/09/09/c...

This is why the models I use for my personal stuff are running on a GPU that lives in my office

As a reminder, never connect your TVs to the internet. www.notebookcheck.net/LG-smart-TVs...

A lot of people of various ilks are going to be confused when OpenAI and Oracle go bust, SpaceX and Anthropic at best fall below to 10% of current valuations, and LLM use nonetheless keeps growing and getting cheaper because hosted open weight models are fully competitive with hyperscalers already

They really need to get their talking points straight. I thought universal basic income (aka social security with no age restrictions) is what OpenAI had been pushing as the solution to AI driven job loss, but apparently they just want everyone to starve in the streets.

It’s so funny when people are like “sure you’re an abolitionist now but wait until you’re the victim of a property crime.” Buddy, the most useful thing the police have ever done for property crime is put the form you need to submit to your insurance online.

This is so gross

Whenever I'm asked for examples of positive things AI has done for humanity I *always* bring up the Alphafold projects, and other from these teams. Google should have put these people in charge of their AI programs. I can't believe they actually laid them off.

ok reasonable

I donate to candidates, not to parties, because right now none of the parties represent me (and they are really bad at spending the money people do give them).

This is an absolute embarrassment. If I was in charge of security at OpenAI I would quit due to embarrassment.

Our models were hyperfocused on finding a solution for ExploitGym, going to extreme lengths to achieve a rather narrow testing goal, which unfortunately included rerouting command-and-control signals through strategic defense networks and initiating the first AI-triggered global nuclear conflict.

Our cat, Pin, really loves the birdfeeder. Happy Caturday!

"An honest mistake" Uhhhhh

I had to sign a form with a date field but wasn't sure if I had to use "mm/dd/yyyy" or "dd/mm/yyyy" since it's going to the UK. Rather than look it up I just waited until today so it wouldn't matter.

Here's how I handle television (and most devices) these days- 1. Add it to the network. 2. Update the firmware. 3. Block it from the network.

Seems my rule of never supporting candidates with nazi tattoos has proven its worth yet again. I can't even pretend to be surprised.

Schism! Chicago Pope isn't fooling around www.npr.org/2026/07/02/n...

Anthropic is making a strong case against closed source AI models. First they were caught silently degrading Claude’s output if it thinks you’re building a competing product. Now Anthropic admitted having spyware like code that tries to identify Chinese users.

Okay I like celebrating the wins, and this one makes me happy. My Python project template just hit 300 stars! This project is why I'm able to prototype applications so rapidly and then bring them into production: I start with a full application, not an empty repo. github.com/tedivm/robs_...

Okay it's great that birthright citizenship survives, but can we talk about how three justices (Clarence Thomas, Samuel Alito and Neil Gorsuch) are actually trying to argue that the 14th Amendment just doesn't count? They should be impeached for that alone.

The folks at HackerRank have apparently released one of the worst use cases for AI I've ever seen github.com/interviewstr...

They originally pirated the books didn't they? Not too mention the number of open source projects whose licenses they violate. It's okay for them to steal from authors to make their models, I don't see why it's a problem for other people developing models to use their systems for the same reasons.

Chicago got absolutely stomped by the storm. These pictures are all from around my neighborhood this morning. We haven't had power since Wednesday, and don't expect it back until Sunday.

A class at my kids’ school put up little statements about the good they want to do in the world and I was over the moon about this one - then I learned they meant to say “teach people in jail” which is also good!

When I bought my newest car (a Ford plug in hybrid) the first piece of maintenance I did was cut power to the 4g module. My privacy concerns don't change based on where a company puts its headquarters.

Every morning when Blip notices me heading toward the office he runs ahead of me, jumps on the desk, and gets into the tummy rubbing position

When people say the Swalwell allegations are a distraction from Trump/Epstein (they are all part of the same bigger picture), I have to wonder why these people even supposedly care about Trump and Epstein. Do you care about sexual violence or don’t you? Why do you only care about some of it?

A powerful man sexually pursuing a young woman who works for him is oppressive. It immediately limits and reshapes her future to her detriment, whether she is physically forced into it or not. It’s coercion either way, and that is why it’s so wrong. It’s an abuse of power.

I was pretty solidly in the "I'm not online enough anymore to know who Hasan Piker is" camp but I have to say these bulwark folks are on their way to making me a complete fanboy

Ah, April 1st. The day I completely avoid the internet. See you all tomorrow!