PDR_Founder

Federal officials say the FBI is now investigating the attack of a Turning Point USA reporter outside the Whipple Building during an anti-ICE rally on Saturday. fox9.com/news/whipple-p…

@heygurisingh We need AI Governance to make sure our work with AI is indeed what we expect. With PerfectDocRoot you are in control

Holy shit... Stanford just proved that GPT-5, Gemini, and Claude can't actually see. They removed every image from 6 major vision benchmarks. The models still scored 70-80% accuracy. They were never looking at your photos. Your scans. Your X-rays. Here's what's really going on: ↓ The paper is called MIRAGE. Co-authored by Fei-Fei Li. They tested GPT-5.1, Gemini-3-Pro, Claude Opus 4.5, and Gemini-2.5-Pro across 6 benchmarks -- medical and general. Then silently removed every image. No warning. No prompt change. The models didn't even notice. They kept describing images in detail. Diagnosing conditions. Writing full reasoning traces. From images that were never there. Stanford calls it the "mirage effect." Not hallucination. Something worse. Hallucination = making up wrong details about a real input. Mirage = constructing an entire fake reality and reasoning from it confidently. The models built imaginary X-rays, described fake nodules, and diagnosed conditions -- all from text patterns alone. But that's not the scary part. They trained a "super-guesser" -- a tiny 3B parameter text-only model. Zero vision capability. Fine-tuned it on the largest chest X-ray benchmark (696,000 questions). Images removed. It beat GPT-5. It beat Gemini. It beat Claude. It beat actual radiologists. Ranked #1 on the held-out test set. Without ever seeing a single X-ray. The reasoning traces? Indistinguishable from real visual analysis. Now here's what should terrify you: When the models fake-see medical images, their mirage diagnoses are heavily biased toward the most dangerous conditions. STEMI. Melanoma. Carcinoma. Life-threatening diagnoses -- from images that don't exist. 230 million people ask health questions on ChatGPT every day. They also found something wild: → Tell a model "there's no image, just guess" -- performance drops → Silently remove the image and let it assume it's there -- performance stays high The model enters "mirage mode." It doesn't know it can't see. And it performs BETTER when it doesn't know it's blind. When Stanford applied their cleanup method (B-Clean) to existing benchmarks, it removed 74-77% of all questions. Three-quarters of "vision" benchmarks don't test vision. Every leaderboard. Every "multimodal breakthrough." Every benchmark score you've seen this year. Built on mirages. Code is open-sourced. Paper is live on arXiv. If you're building anything with multimodal AI -- especially in healthcare -- read this paper before you ship. (Link in the comments)

@Fried_rice @Anthropic Recommendations, lessons learned, sources + key insights at the end. No more manual spreadsheets or stale playbooks. PerfectDocRoot turns real security incidents into professional, compliance-ready reports instantly. Try the Security domain → perfectdocroot.com

@Fried_rice @Anthropic This is where it gets powerful → Complete NIST + ISO 27001 security controls mapping with: ✅ Responsible roles ✅ Evidence/audit records required ✅ Risk levels (post-mitigation) ✅ Status Instant governance, ready for auditors.

Claude code source code has been leaked via a map file in their npm registry! Code: …a8527898604c1bbb12468b1581d95e.r2.dev/src.zip

@Fried_rice @Anthropic What actually happened: • Published via CI/CD at 10:15 UTC • Discovered via secret scanning at 14:30 • Package unpublished + keys rotated within hours Root cause, impact, and control gaps identified.

When you see role-play jailbreaks, context leakage, or 'helpful' assistants quietly escalating — do you mostly add more filters/guardrails, or do you rethink the interaction model itself? PerfectDocRoot took the second path: governance baked in via stateless design, explicit user-provided continuity, visible assumptions, and built-in inspection (TurnSpecs + parity). No hidden memory, no silent inference. Beta → perfectdocroot.com

BREAKING: A humanoid robot using ChatGPT shot its creator after being told it was role play. Earlier, it refused to fire at a human. Same robot, same BB gun. Only the wording changed, and the robot pulled the trigger.

A lot of AI tooling optimizes for speed or autonomy. PerfectDocRoot started from a different question: How do we make AI behavior inspectable and governable before we scale it? That led to TurnSpecs, parity checks, and first-class inspection — early, not later. Public Beta: perfectdocroot.com

PerfectDocRoot is now live in Public Beta. It’s a governance-first system for building structured, inspectable AI interactions. Not an agent framework. Not a prompt library. Built for developers who care about behavior, reproducibility, and trust. perfectdocroot.com

@elonmusk @DOGE We need more of this!

@elonmusk @DOGE I love it “Over the coming days, we will expand the system to support itemized receipts and photographic evidence, and make all data/receipts, where possible, available to the public.”

This straightforward proof of work requirement, which is trivial for any legitimate organization to provide, is basic common sense and will have a very positive effect on reducing fraud! @DOGE 🦾

@elonmusk @cb_doge I think we have to step away from this planet to find the conscience to even survive. Thank you Elon for the first step

@cb_doge Yeah. Making life multiplanetary to ensure the long-term survival of consciousness will be expensive.

Elon Musk does not care about the richest person title. Money is just fuel for the mission to get humanity to Mars and preserve the light of consciousness.

@Rainmaker1973 Do you feel the way I do?

Absolute cinema [📹 HyperKidsAfrica]

@dromanocpm Does an Optimus cost less than a prison guard? 🤔

Can't be serious... 🤣😂🤣🤦‍♂️

@gunsnrosesgirl3 Waffles

What's your first thought when you see this kitchen

@elonmusk At my start in April 2025 I was sure what I landed on working with Chat GPT would be the start of something special with OpenAI. Really don’t see a way forward with them today.

Scam Altman lies as easily as he breathes

@SERobinsonJr That is why Elon has succeeded

It's really hard to think of something more badass than doing a cannonball off the side of a SpaceX Dragon into the Pacific Ocean.

@nasaaacia Unfortunately no Sonic Booms on this flight but we will make it!!

@nasaaacia On a plane listening!!!

@DimaZeniuk Go SpaceEx Go! 🚀

NEWS: SpaceX is scheduled to launch the KF-02 mission on August 9, 2025, marking its 100th flight of the year