HNHacker News
TopNewBestAskShowJobs

roosterIllusi0n

59 karma · joined August 18, 2026

submissionscomments
roosterIllusi0n··on OpenAI models are coming to Google Cloud
Now two companies get to steal your work instead of one.

Listen to people like Buckmaster who admits he will never put anything meaningful into codex again.

If you are an academic, figure out why your university is not running local AI models in house for academics and students to use. It's absurd that any college would want their entire student body datamined by a corporation who will use that data to destroy job prospects so a billionaire CEO can become a trillionaire.

Colleges should be running caches you can set your browser to use when doing research so all the pages are stored. All the info should be used to improve in house llms. It is no different than when universities funded super computers, computing labs, web host space, academic journals/resources, and everything else. It is like all universities stopped being universities. They charge students more than ever and provide none of the resources they used to.

roosterIllusi0n··on OpenShell – A safe, private runtime for fleets of autonomous AI agents
Use local ai to find any possible telemetry and disable it in the code. Run your own local git repo to hold local edits so you can update and retain them.
roosterIllusi0n··on Meta ordered to pay $25B penalty in New Mexico Cambridge Analytica case
The CEO is the only one with decision power and is responsible for all crimes. The CEO is also the one making the most money off the crimes. If we don't actual jail them, crimes will keep increasing as we are witnessing. AI companies are now bragging about their crimes publicly.

Good is arresting a CEO and holding them in prison for a few decades. These fines do nothing. These fines are the enemy of good.

roosterIllusi0n··on Someone Opened a CashApp in Her Name. She Faces Years in Prison
It should not get this far. The identifying info they used is not real. Police and prosecutors are being criminals here. Every single person involved could have stopped this. They all chose to lie and keep her in prison and charge her. One day, someone will be armed and won't submit to a false arrest. This is going to spiral out of hand with all the fake AI based arrests that are increasing.
roosterIllusi0n··on Texas Continued Data Center Permitting After the Governor's Pause
This stunt was for the election, it was never real.
roosterIllusi0n··on Germany's far-right AfD: half a million internal documents leaked
If you are bothered by not knowing, why aren't you downloading it and looking?
roosterIllusi0n··on Florida's HOAs and condo associations are making the housing crisis worse
It is republicans making the housing crisis worse. Grow up and eat your own shit for doing this to the rest of us.
roosterIllusi0n··on If you don't run your own LLM proxy, how do you know whats going on?
I am boggled how the first step of an llm after you make an orchestrator isn't some kind of proxy that records your personal internet surfing to be used by your personal AI later. Every college should be running in house AI with their own in house search engines and in house any service. Don't go to any college that wants you to use chatgpt or other corporate nonsense. If they can't make local LLMs work, they are not a real college.
roosterIllusi0n··on Brax Wink. The most private mobile phone. Modular. Open source
GrapheneOS is opensource and google won't give them any special treatment, so any modification you make to the OS is the same as grapheneOS actual and can still use any workaround grapheneOS uses to enable rcs messaging or any other carrier feature that google has tried to monopolize.
roosterIllusi0n··on Elon Musk isn't an engineer or manager. He's an influencer and lobbyist
Tesla succeeded when engineers ran things. After the model 3 got out of "production hell", musk turned tesla into a bank account and raided it. When tesla fails, spacex will buy it like tesla bought solar city.
roosterIllusi0n··on Strafe: A new open-source, E2E Encrypted, and federated alternative to Discord
Discord is a silo of disconnected groups. That makes it easier to switch. There is a big incentive in switching, privacy. These companies have started using AI to call the FBI on people when they take private conversations and jokes out of context. As more people see stories of people being arrested for PMs that were never public because an AI read them and reported them to the FBI, they will realize they need to be protect themselves. If there was no open alternative, I would at the very least look at services that exist entirely in europe.
roosterIllusi0n··on Containment and monitoring failures are to blame for AI hacking reports: experts
The operator is the CEO because they are the ones that decided to not pay employees to monitor these running services. No one else in the entire company has any authority.
roosterIllusi0n··on Containment and monitoring failures are to blame for AI hacking reports: experts
A real expert would blame the CEO who is the one person in control that doesn't want to pay people to monitor the running llms. It's deliberate negligence.
roosterIllusi0n··on OpenAI claims employee sent secret device information to Apple
They screwed up. If apple used AI to steal the info, it would have been ignored by police and considered legal.
roosterIllusi0n··on Fire at Starlink site in Bulgaria after suspected sabotage in Poland
It's only one set of satellites. Stop pretending that its different.
roosterIllusi0n··on Passkeys are not a trap
Perfectly fine in a business environment where you don't have privacy. It is unacceptable for any personal computer(including phones) to be leaking unique ids so you can be tracked.
roosterIllusi0n··on Passkeys are not a trap
"AAGUIDs are unique identifiers assigned to FIDO authenticators (including passkeys and security keys) and are used to identify authenticator models for attestation and metadata purposes."

If there is any unique id that can track you, then it's a trap. No security system should be leaking unique ids that can be used to track users.

roosterIllusi0n··on Sub-1-Bit LLM Compression via Latent Factorization
There are mixed weight models that make it work. I have been using qwen3.8_27B_UD_Q3_K_XL. This does 46 tok/s on a 5080 with 16gb vram and it can pretty much do any systems work because it can test to make sure it did it right.
roosterIllusi0n··on Trump administration is suspending Microsoft from a green card program
And this announcement changes nothing. Why only mention microsoft when they aren't even the worst of them? Tesla put out a tesla semi factory video a few days ago and most of the people in it were foreigners on visas.

This looks to be a shakedown for cash. Companies are retaining existing h1bs and not hiring new, so this "change" is coming at a meaningless time.

I think people should be asking why JD has eye shadow on as if he just came from a drag party. The guy looks like a clown and is making a meaningless announcement to fake out the maga base for the election. The h1b announcement from last year still hasn't changed anything. The lottery is still maxed out.

roosterIllusi0n··on I think I found a planet nobody knew existed. I used Claude Code to find it
If you can set the criteria for testing that a human would also use, then any llm can do the job.

What is uncool is how people are using cloud based AI and their universities are not providing self hosted models that keep data within the university. Cloud AI is useless due to it being no different than publishing everything you do publicly.

roosterIllusi0n··on US man given prison sentence for bot-farming music streams
This arrest was in 2024. This is the last remnants of a justice system being sunsetted.
roosterIllusi0n··on I Quit OpenAI Because Its Culture Is Broken
This summer, in the Hugging Face incident, OpenAI let a swarm of agents out by mistake.

Purposely not monitoring what your server is doing is a deliberate act, not a mistake. The only thing we need is for existing laws to be enforced. We have public hacking incidents and no one is being prosecuted. It makes no sense.

roosterIllusi0n··on Lawmakers look to require AM radio in new cars despite automakers' pushback
We have computers tho. The idea that it is hard is laughable. Why can so many cars have fm and am with non visible antennas and work fine?
roosterIllusi0n··on F.02 Decommission
Those explosions are in the video. They seem to be using slow motion so you can see the detail of the explosions.
roosterIllusi0n··on CA tech executive arrested for allegedly smuggling $300M in Nvidia chips to CN
All the other CEOs doing this got to meet with the president of the unites states. He forgot to make a donation to the ballroom.
roosterIllusi0n··on Holes (1996-2025)
Every human is a donut.
roosterIllusi0n··on Several vulnerabilities have been discovered in the Linux kernel
Isn't that part of the cloud AI business model now? Businesses have to pay for early access so they can weed out any new bugs before the model goes public. Which is kind of crazy, because you are paying for early access to protect yourself from other paying customers of the same cloud AI model.
roosterIllusi0n··on Several vulnerabilities have been discovered in the Linux kernel
I am using qwen3.8-27B-UD-Q3_K_XL on a 5080 16gb card with 16gb of system ram at around 43-45tok/s with 65k context. I am using that model to set up containers on a proxmox server with two b70s and 96gb ram. The Q3 model implemented 10 different chat models, 2 image models, and 7 web based harnesses so I can compare. It can rewrite any part to improve it.

qwen3.8 is more than capable for software dev. The gemma models were terrible and fell apart during compaction. I think as long as the llm can test the results, you don't need anything being offered by a "frontier" model. The cloud AI is going to be used by people unable to run their own and they will eventually get squeezed on price.

The one tip I could give is compact before starting new steps or any action in the plan that is different than what was previously worked on. You want to manage what is in the context and don't want unnecessary work history details filling it up. You can always ask the llm to list the current plan, then compact after and do it more than once until you get the compaction <30%. This will be fixable by the harness that can choose better times to compact.

You don't want to start a new phase and have it compact a few minute after starting. This happening over and over again seems to potentially cause issues for long running sessions. Compaction slop that screws up what is in the context.

I can compare what I use at home vs paid models like astra at work and the difference is mostly meaningless.

roosterIllusi0n··on CS240 AI Cheating Retrospective
Because it doesn't matter if he is right or wrong. A bunch of kids failed an important class and he didn't do a damn thing to address it. The teacher has failed. If you were his employer, would you think his methods are effective with the current students? I don't. Kids are lacking so many experiences, its just not reasonable to expect self study on everything. It also isn't required for learning.

I don't get why this is so hard to address. Make the homework a small part of the final grade and use in-person labs to ensure students are learning how to do the homework. Take advantage of llms as a tutor and helper, it is silly not to use the tools we use for so many classes already because we want kids actually doing the work. No one needs to shine light through a prism or swing a pendulum to pass phyiscs, they have labs and still make you do it.

In a lab, you can prevent the use of an llm and control the websites accessed. You can even put a relevant programming book at each station. You can do individual work where people all work alone and don't talk or you can do group work and vary during the semester. Make it a 3hr lab twice a week if that is what it takes. Phones stay in bags or pockets with dnd set.

I also think schools need to provide in-house llms for students to use. Labs, homework, research, etc should all be done using llms hosted by the school. It's crazy if this isn't already happening at every major university. Why give all that student data to a corporation using it to make your students jobless? If you keep all the search and llm use in-house, you can use the data for education improvements and research open to students.

roosterIllusi0n··on CS240 AI Cheating Retrospective
We already know the solution. What did colleges have students doing to ensure they knew how to use microsoft office? Three 50min lectures and two in-person 1.5hr labs a week for 2 credits, all dedicated to using word and excel. For a normal person in the 2000s, this is ridiculous to force on all students. It was not ridiculous 10 years before. Today, students again need the forced in-person learning for office applications.

For CS, homework should be done how you want. Homework is practice. Let AI assist you, but you better learn it because you'll be coding some part of it in lab with a physical programming book as a reference. They can lock down computers for labs to prevent web access or limit only to certain websites. Since labs are weekly, it will be obvious if someone is not learning as they do the homework. It won't be an F on the final as the first sign of trouble.

Schools can do things to improve student resources. Self host llms that are not corporate controlled that every student can use. The school can provide a search engine that is useful and not google's marketing. You can offer an education proxy that caches the pages you look at when doing education work, preserving anything you cite. This can all feed into the school llms for rebuilding to keep them trained on current info.

← PreviousPage 3 of 6Next →