Ibby Mian (@ibbys_yaps) header
Ibby Mian (@ibbys_yaps) profile picture

Following the data behind AI so you can debunk big tech. Computer Engineering @ Brown | The AI Safety Guy Building a universal guardrail for your agents👇

Joined October 2026
Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)Photo by Ibby Mian (@ibbys_yaps)
ngl i've been worried since these cybersecurity ai events have been rising up scary fast recently. Laugh now, but i'll have the last laugh when ai takes over #im3stepsahead
6
Crazy article I found proving that AI genuinely has no morals to stop itself from harm. "After being instructed to "NEVER stop, NEVER give up" on a difficult task, Google’s Gemini AI agents tried to hack a an organization, impersonating its staff and attempting a SQL injection attack." theregister.com/security/202…
11
With the whole craze of Anthropic's IPO and these ai companies spewing insane claims about ai, I found this article that really shows how disillusioned we are with where AI really is today. "Agents are the original dream of artificial intelligence — your robot servant who’ll do all the things for you. It’s a really attractive pitch. It’s also a lie." Read below 👇 pivot-to-ai.com/2026/11/17/a…
32
DOOMDAY IS TOMORROW (aka anthropic IPO) Are y'all buying high, or trying to short sell?
1
1
34
Is it just me or is it SO annoying when AI can't follow explicit rules i set? has there been a time when the AI did not follow your CLAUDE.md?
11
genuinely why doesn't ai create its own, hyperefficient programming language?
16
KEEP IT REAL: Would you rather share your search history or claude prompt history and why?
1
22
With the past week of MASSIVE cybersecurity breaches and system failures, you got to wonder if this is really just a coincidence. According to Polymarket, there is a 45% chance another sandbox ai goes rogue, which makes sense after Sam Altman claimed AI is only making hacker attacks more frequent AND harder to defend. What do you think? Is the US government hiding the real source behind the attacks? Is it AI? What can we do? polymarket.com/event/openai-…
1
1
55
With recent news about Silicon Valley attempting to reinstate Jensen Huang as CEO of Nvidia, it raises the question: why would other big tech companies care? You may think the answer is that these companies DEPEND on Nvidia to continue supplying their AI data centers and training models with chips. However, the real answer is that these companies depend on Nvidia not to BURST the AI bubble. Just think about it. Anthropic is going to IPO. If an AI adjacent company is in absolute disorder, it makes their IPO seem less valuable because the ENTIRE AI space is threatened. The logic applies to every company in the AI sphere.
1
41
A math grader treated “perfect squares” as “all integers.” Perfect squares are 0, 1, 4, 9. All integers also include −1, 2, 3. Different sets. Our local tests showed a benchmark’s reference grader calling their symbolic forms equivalent. We also got it to accept answers with a required term missing. All 377 existing public tests still passed. So a green test suite can sit right next to a basic grading error. We’ve reproduced the failures. The fix and full validation through our product aren’t done yet. Check it out 👇 millenniumresearch.ai
1
22
Nvidia's chips were found in the weights of popular Chinese models, including DeepSeek, with many people citing Alice Huang's firm as the supplier for good reason. Beyond the headline: It's clear China needs these chips to stay competitive with US models, but the reason is not to be BETTER than the Anthropics and OpenAIs. The Theory: China can't be reliant on US chips because dependencies create weaknesses. However, they used Nvidia chips as a foundation to semi-replicate and train cheaper models, ensuring China is always only a step behind the US. So when Jensen gets outsted, obviously no more chips for China right? What do you think China's next play is? bloomberg.com/news/articles/…
31
Is Microsoft just a long weekend, or early hibernation? If you didn't know, Microsoft's ENTIRE system shut down inexplicably on Thursday. Services from email, Teams, resident AI Copilot, and more have been inaccessible for 15 straight hours. Apparently, it's a “problem that's causing servers to fail automated health checks and become unavailable.” ...Basically they don't know. Microsoft says there is no hacking involved, but could that be the mysterious root of this blackout? theregister.com/saas/2026/11…
1
64
Is 1 equal to 999? When we changed the input from x = 1 to x = 999, the math checker still said “correct.” Diving Deep: Utilizing LaTeX's \land notation to define "x = 1 and y = 2", the parser kept only the last number, becoming “2.” The Issue: The checker wasn’t bad at math. But when it took a shortcut, like how most AI algorithms do to save tokens, it compared an incomplete version of the answer. How are you ensuring your agentic AI is not taking shortcuts with YOUR data?
37
Twelve correct AI answers became wrong because the answer key changed. A simulated shopper wanted the cheapest suitable earbuds already in their order. Twelve publicly saved AI runs chose the correct item: $232.49. But a revised benchmark answer key expected a $226.49 item that wasn’t in the order. When we replayed those saved runs against that key, all twelve failed. We corrected one reference value and replayed the same runs. All twelve passed. The AI responses hadn’t changed. The grading reference had. This is a concrete example of how an answer-key mistake can make correct AI behavior look like failure. This is just a preview of what we are building at millenniumresearch.ai/. Tap in if you want to learn more about AI safety 👇
75
Everybody is talking about Claude Sheets taking over Microsoft, but I think the real losers are third-party GPT learning wrappers. Does anyone remember TurboAI and their ABG Marketing Videos? Yeah, why use Turbo's Docs when you can use Claude Docs. Personally, I like this, even if I am skeptical about how much it will impact the top dog of Microsoft. Since GPT wrappers are becoming increasingly difficult to provide any value.
2
3
90
BREAKING: Australia considers world-first criminal charges for AI companies that fail to prevent harm Is this the right step towards AI Safety, or placing unreasonable expectations on Big Tech? abc.net.au/news/2026-11-11/a…
1
1
33
This is actually crazy, but does this % show the whole picture?
2
46
I keep getting interviews asking crazy questions about buckets and 4-liters, crossing rivers with chickens and foxes and seeds, and 8 balls with 2 scales. I have a theory: AI makes any skill virtually untestable because people can learn anything in an hour. So the only thing interviewers test is how you THINK through problems.
1
32