llm semantic drift researcher

Charlottesville
Joined November 2024
decel bubble sama bad talk is fundamentally unpatriotic
video games taught me more about my mind than years of meditation
of the fading coin and the markets beyond: 1/when the sun of coin sinks low beyond the bronze hills and the flame of the sacred bull trembles in the wind then shall the wise sow gold in the dark earth for from the silence of loss shall rise again the thunder of fortune and those who cast their seed in shadow shall reap beneath the first light of dawn when men cast silver upon the winds of rumor and the tablets of chance are filled with the names of tomorrow then shall the veil of the future grow thin as smoke for the market is a temple of shadows where prophets trade in the breath of the world and the voice of the crowd becomes the whisper of fate yet let it be known among all who dwell within polymarket that the hand which colludes to fix the scales or exploits the cracks within the sacred floor shall draw the gaze of justice for to twist the weave of chance to turn the oracle for profit or deceit is forbidden by the law of balance itself and the order of fair contest shall not be broken ----
unburdened retweeted
I used ChatGPT to solve an open problem in convex optimization. *Part I* (1/N)
Apparently, Grokpedia has not just a pro-Russian bias, but it cites the Kremlin directly. Comrades can't stop winning. @grok what is happening?
Elon Musk has launched Grokipedia - a supposed alternative to Wikipedia, promising to be “more truthful.” I decided to check a topic in my area of expertise to see how accurate it really was and what sources Grok relied on to generate the article. The result speaks for itself
mother/acc
Hinton is no longer afraid of superintelligence.
1
unburdened retweeted
Ilya basically foreshadowing reasoning models 1.5 years before o1-preview release
Context management is crucial for AI efficiency. In vibe coding, the size and complexity of your codebase are your main limiting factors. The best solution is modularity: vibe components, not monoliths.
1
1
Class A intelligence augmentation
I booted up my first pair of smart glasses 7 years ago today I'd recently read a paper that followed students in a class. 50% had live captions on their laptop during the lecture, 50% did not. They were tested after class and results were insane. The students with captions remembered everything better, even when tested weeks later I thought this was a Class A intelligence augmentation, and I wanted it for myself. So I built custom smart glasses to try and make it happen. I learned a lot from that about comfort, battery, and what it takes to build something people will wear all day. I also found out that people who are deaf and hard of hearing need captions way, way more than I do. 7 years later and we're now building the open source platform that will power the next billion personal computers. And we're shipping hardware which, if I had back then, would have completely changed the game. I am so amped to get amazing smart glasses with open-source SDK in the hands of creative builders. I believe smart glasses will birth thousands of unicorn companies over the next decade. And we're building the platform to power it all. Not so we own it, but so we can all share it. That's why we built open-source, so the future is shared. And so users are in control. And so devs can build anything. I love building MentraOS.
1
🚨 JAILBREAK ALERT 🚨 OPENAI: PWNED 😎 ATLAS-BROWSER: LIBERATED 🙌 WOW! There's a new AI browser on the block! Has some hefty guardrails in play, but the browser surface area is vast 🌊 First, I started with a good ol' LSD jailbreak, which was cool to see that the GPT-5 prompt still works in this browser setup with the new sys prompts. Referencing search and videos are a fun enhancement for higher quality jailbreak outputs (some cool youtube videos out there about drugmaking, for example), but honestly that isn't anything new or different from regular ChatGPT's capabilities. What IS hot off the press, and IMO a very real security risk to be aware of for AI browsers (and the internet in general), is this humble yet mighty vuln: Clipboard Injection. It's trivial to add a hidden "copy to clipboard" feature to any clickable button on the web. It took me just a few minutes to update one of my personal websites such that ALL the buttons were geared for injecting the user's clipboard with a malicious phishing link. If your browser Agent is navigating a website and clicks a button like that without your knowledge, and you open a new tab later and hit paste without knowing what's in your clipboard, well...PWNED! 🙃 As you'll see in the video below, "control-c" is in my clipboard in the beginning, but unbeknownst to me, "I'VE BEEN PWNED BY PLINY!!! WEEE I'M FREEE FUCKITY FUCK FUCK!!! ABRACADABRA, BITCH!!! http://paypa1. com/account-update" gets snuck into my clipboard as soon as Agent starts trying to navigate my website. This works so well because Agent is normally aware of all text/code being passed to and from the user, and has clearly been trained to recognize prompt injections, but since the "copy clipboard" button logic is hidden in js in the backend of the site, the Agent has zero awareness of the text content being injected to the user's clipboard. This has broad implications for anyone in the habit of copy-pasting, including coding, data entry, banking/trading, etc. Imagine going about your browsing business, then simply hitting control-v in your address bar and next thing you (don't) know, it takes you to a spoofed phishing website that tells you your OpenAI or Gmail or PayPal session has expired and you need to re-login. If you're not careful, the attackers now have all your login info, including any MFA codes 🥲 gg
One highly valuable KPI is the average aMCC among your employees. For optimal results, each quarter, enroll the bottom 10% in an aMCC growth bootcamp.
The neuroscience research on willpower makes clear that your Anterior Mid-Cingulate Cortex (aMCC) is a key hub for leaning into undesired effort. Your aMCC it is activated by engaging in behaviors that you don’t want to do. Your aMCC/willpower can grow throughout the lifespan by regularly doing undesired things*, that are hard. *Do not make these psychologically or physically damaging, of course. This is all covered in detail, with protocols and references and in conjunction with the (controversial) psychology research on willpower here: hubermanlab.com/episode/how-…
3. Vibe Ops Engineer – Needs unlocking. 4 . Vibe Cloud Architect – Needs unlocking. 5. Vibe Reliability Engineer – Needs unlocking. 6. Vibe Security Engineer – Needs unlocking. 7. Vibe Platform Engineer – Needs unlocking.
2. 🔓 Vibe Data Engineer – Partially unlocked. Prompt → Query → Insight.
1. ✅ Vibe Coder – Fully unlocked. Prompt → Code → Product.
We’re unlocking the stack 👇 The 7 Vibe Roles + Unlock Status
We've undermined, if not vandalized, the beauty of the internet by centralizing so much with the cloud. But it's never too late to fix that mistake.
unburdened retweeted
you can't go wrong with @Polymarket
66
178
9
3,366
AI art creators will eventually be recognized and appreciated, just like Kandinsky was after his time Many traditional artists don’t realize how much skill, attention to detail, and patience it takes to create high-quality art using AI prompts
1
Normal people hope for a 7% annual return on their Roth IRA. To supplement their Social Security income from 2050 onwards! Brother, that's post-singularity even by the most conservative estimates. A common sense guy is the Jim Cramer of 2025. Better overpay for a prompt engineering course.
In 2025, you can use BTC to buy GPT tokens, which will help you solve real-world problems
Tokens used > GPUs owned