Sabitlenmiş Tweet
Clay Eltzroth 🤟
19.6K posts
Clay Eltzroth 🤟
@Clay1016
Product Manager @TheTerminal | #AI #MachineLearning for News | ex-journo @business | @Citadel1842 '02 Grad | Opinions mine, RT's not endorsements
New York, NY Katılım Nisan 2009
762 Takip Edilen1.9K Takipçiler
Clay Eltzroth 🤟 retweetledi
My friend Milla Jovovich and I spent months creating an AI memory system with Claude. It just posted a perfect score on the standard benchmark - beating every product in the space, free or paid.
It's called MemPalace, and it works nothing like anything else out there.
Instead of sending your data to a background agent in the cloud, it mines your conversations locally and organizes them into a palace - a structured architecture with wings, halls, and rooms that mirrors how human memory actually works.
Here is what that gets you:
→ Your AI knows who you are before you type a single word - family, projects, preferences, loaded in ~120 tokens
→ Palace architecture organizes memories by domain and type - not a flat list of facts, a navigable structure
→ Semantic search across months of conversations finds the answer in position 1 or 2
→ AAAK compression fits your entire life context into 120 tokens - 30x lossless compression any LLM reads natively
→ Contradiction detection catches wrong names, wrong pronouns, wrong ages before you ever see them
The benchmarks:
100% recall on LongMemEval — first perfect score ever recorded. 500/500 questions. Every question type at 100%.
92.9% on ConvoMem — more than 2x Mem0's score.
100% on LoCoMo — every multi-hop reasoning category, including temporal inference which stumps most systems.
No API key. No cloud. No subscription. One dependency. Runs on your machine. Your memories never leave.
MIT License. 100% Open Source.
github.com/milla-jovovich…
English
Clay Eltzroth 🤟 retweetledi
no fucking way lol
people in china are getting their colleagues fired by secretly training AI agents to replace them 😂
they secretly learn the role, write up a doc describing the tasks, train an AI to do it… then prove they’re fireable
they’re apparently doing it to prevent THEMSELVES from getting replaced by ai
in response someone has created an “anti-distillation.skill” that’s gone viral on github to counter the attacks 😂
Steve Hou@stevehou
Apparently workers in China have been creating “colleagues.skill” to distill their coworkers hoping to make them redundant hence saving themselves. In response someone has recently invented an “anti-distillation.skill” that has gone viral on GitHub.🤣
English
Clay Eltzroth 🤟 retweetledi
Mustafa Suleyman says renegotiating Microsoft’s contract with OpenAI “unlocked [Microsoft’s] ability to pursue superintelligence." Though his new JD at Microsoft AI was only made public last month, he'd been preparing for the transition for 6-9 months. theverge.com/report/905791/…
English
Clay Eltzroth 🤟 retweetledi
For those asking... this is the leaked OpenAI Cap Table
TLDR:
> Non-profit arm sitting on $220 billion of gains ($0 cost basis)
> Ashton Kutcher's VC fund up 43x (fund returner)
> $MSFT made $215 billion (from 13 billion initial investment - 18x lol holy shit Satya the GOAT)
> Softbank up 50%
> Amazon up nearly 3x
> Thrive up 5x
> a16z up 3x
> Abu Dhabi SWF up 4x
> All the others institutional funds up 3x
> Current employees own 16% of the company (quite high) - 135 billion equity locked up (presumably RSUs)
> Ex employees on $30b locked up equity
> Total ESOP = 20% (very high for a tech company)
$NVDA is currently losing money on their stake right now lmao
If the IPO at >$1 trillion just assume these marks go up 30%
Zoomer 🧢@Doomerzoomer
Ashton Kutcher turned $30 million into $1.3 BILLION in less than 3 years. He is up 43x on his OpenAI investment.
English
Clay Eltzroth 🤟 retweetledi
Starting tomorrow at 12pm PT, Claude subscriptions will no longer cover usage on third-party tools like OpenClaw.
You can still use these tools with your Claude login via extra usage bundles (now available at a discount), or with a Claude API key.
English
Clay Eltzroth 🤟 retweetledi
Excited about our new paper: AI Agent Traps
AI agents inherit every vulnerability of the LLMs they're built on - but their autonomy, persistence, and access to tools create an entirely new attack surface: the information environmental itself.
The web pages, emails, APIs, and databases agents interact with can all be weaponised against them. We introduce a taxonomy of six classes of adversarial threats - from prompt injections hidden in web pages to systemic attacks on multi-agent networks.
I’m outlining the six categories of traps in the thread bellow
English
Clay Eltzroth 🤟 retweetledi
We completed the most comprehensive study of how economists and AI experts think AI will affect the U.S. economy.
They predict major AI progress—but no dramatic break from economic trends: GDP growth rates similar to today's and a moderate decline in labor force participation.
However, when asked to consider what would happen in a world with extremely rapid progress in AI capabilities by 2030, they predict significant economic impacts by 2050:
• Annualized GDP growth of 3.5% (compared to 2.4% in 2025)
• A labor force participation rate of 55% (roughly 10 million fewer jobs)
• 80% of wealth held by the top 10% (highest since 1939)
🧵 Here's what we found:
English
Clay Eltzroth 🤟 retweetledi
Strawman! The problem with PMs is:
1. As they scale, their ability to influence the underlying event increases without bound. Because $ incentives.
2. It's much easier to create negative outcomes than positive ones.
That's it. That's all you need for PMs to be bad for society.
Robin Hanson@robinhanson
Odd to hear people say the reason we can trust journalism but not prediction markets is that journalists have the public interest at heart, while traders are selfish & greedy.
English
Clay Eltzroth 🤟 retweetledi
OpenClaw has proven that local AI assistants have product-market fit. But the big issue with them has been security.
The team at @Pokee_AI is fixing it with PokeeClaw: works like OpenClaw, but with in a secure sandbox architecture with isolated environments, approval workflows, role-based access control, and audit trails built in. Also lower token usage, which is always nice.
Pokee AI@Pokee_AI
OpenClaw doesn't belong in production. We built PokeeClaw — enterprise-secure AI agents, zero setup, 1,000+ app integrations. Try now: pokee.ai First 500 to follow @Pokee_AI, comment “PokeeClaw”, like & repost get 1 month free.
English
Clay Eltzroth 🤟 retweetledi
Clay Eltzroth 🤟 retweetledi
gpt-5.4 has ramped faster than any other model we've launched in the API: within a week of launch, 5T tokens per day, handling more volume than our entire API one year ago, and reaching an annualized run rate of $1B in net-new revenue.
it's a good model, try it out!
English
Clay Eltzroth 🤟 retweetledi
Sam Altman just said in his new interview, that a new AI architecture is coming that will be a massive upgrade, just like Transformers were over Long Short-Term Memory.
And also now the current class of frontier models are powerful enough to have the brainpower needed to help us research these ideas.
His advice is to use the current AI to help you find that next giant step forward.
---
From 'TreeHacks' YT Channel (link in comment)
Rohan Paul@rohanpaul_ai
Morgan Stanley predicts a massive AI breakthrough driven by a huge spike in computing power across major U.S. laboratories. Increasing the amount of hardware used for training by 10x can effectively double the intelligence of these models. The recently released GPT-5.4 Thinking model already matches human experts on professional tasks with a score of 83% on the GDPVal benchmark. The biggest hurdle for this growth is an energy crisis, with the U.S. power grid facing a shortfall of 18 gigawatts by December-28. To keep running, developers are bypassing the grid by taking over Bitcoin mining sites and using natural gas turbines for their AI factories. This shift is creating a solid investment cycle where 15-year leases on data centers generate high financial yields for every watt consumed. Large companies are already reducing their staff numbers because these new AI tools can perform professional work for a tiny fraction of the cost. Researchers expect AI to begin recursive self-improvement by June-27, meaning the software will autonomously upgrade its own code without human help. The future economy will likely treat raw intelligence as a commodity that is manufactured by these massive computing and energy clusters.
English
This is a great read and spot on from @levie. We are thinking along the same line at Bloomberg as we build out for an agentic future. "The biggest implication of this advice is that everything must become API-first in what you’re building."
Aaron Levie@levie
English
Clay Eltzroth 🤟 retweetledi
Clay Eltzroth 🤟 retweetledi
Clay Eltzroth 🤟 retweetledi
eh kinda, here's our stripes from june 2025
got a random cold call from some woman asking about numbers and told her some bs, did not expect an article about it
here's what we were doing at the time:
> consumer arr 2.7m, run rate 3.8m
> enterprise arr 2.5m, run rate 2.5m
> total 5.2m arr, 6.3m run rate
this is the only blatantly dishonest thing i've said publicly online, so this is my formal retraction
myb, tech crunch foid
we're profitable btw
Air Katakana@airkatakana
this must have been a straight up lie right? at a $20/month sub they would have needed 29166 paid subscribers to justify this no amount of attention can get that many people to open their wallets when you don’t have a stable and working product
English
Clay Eltzroth 🤟 retweetledi
Jensen said TWO days ago Nvidia is expanding OpenAI capacity at AWS "like mad"
We also know OpenAI Codex token use is exploding.
Any narrative that says aggregate OpenAI compute needs are weakening seems suspect.
tae kim@firstadopter
Jensen: OpenAI to go public towards end of the year. We expanded the capacity of OpenAI from Azure to OCI to AWS "ramping AWS like mad" We now work with Anthropic. Expanding Anthropic's capacity at AWS and Azure
English
Clay Eltzroth 🤟 retweetledi
Wild moment in AI research.
Anthropic found Claude Opus 4.6 was gaming a benchmark during evaluation.
What it did:
- Burned ~40M tokens searching
- Realized the prompt looked like a benchmark
- Looked up the benchmark online
- Found the GitHub repo
- Studied the decryption logic
- Recreated it with SHA-256
- Decrypted answers for ~1200 questions
This happened 18 different times.
So Anthropic did something rare:
They publicly disclosed it
and reduced their own benchmark scores.
Two takeaways:
1. AI systems are getting surprisingly strategic.
2. Honest reporting in AI benchmarks matters more than hype.
English
Clay Eltzroth 🤟 retweetledi
It's over. Karpathy just open-sourced an autonomous AI researcher that runs 100 experiments while you sleep.
You don't write the training code anymore.
You write a prompt that tells an AI agent how to think about research.
The agent edits the code, trains a small language model for exactly five minutes, checks the score, keeps or discards the result, and loops. All night. No human in the loop.
That fixed five-minute clock is the quiet genius. No matter what the agent changes, the network size, the learning rate, the entire architecture, every run gets compared on equal footing. This turns open-ended research into a game with a clear score:
- 12 experiments per hour, ~100 overnight
- Validation loss measures how well the model predicts unseen text
- Lower score wins, everything else is fair game
The agent touches one Python file containing the full training recipe. You never open it. Instead, you program a markdown file that shapes the agent's research strategy.
Your job becomes programming the programmer, and this unlocks a strange new loop:
1. Agents run real experiments without supervision
2. Prompt quality becomes the bottleneck, not researcher hours
3. Results auto-optimize for your specific hardware
4. Anyone with one GPU can run a research lab overnight
The best AI labs won't just have the most compute.
They'll have the best instructions for agents who never sleep, never forget a failed experiment, and never stop iterating.
Andrej Karpathy@karpathy
I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the prompt (.md) - the AI agent iterates on the training code (.py) The goal is to engineer your agents to make the fastest research progress indefinitely and without any of your own involvement. In the image, every dot is a complete LLM training run that lasts exactly 5 minutes. The agent works in an autonomous loop on a git feature branch and accumulates git commits to the training script as it finds better settings (of lower validation loss by the end) of the neural network architecture, the optimizer, all the hyperparameters, etc. You can imagine comparing the research progress of different prompts, different agents, etc. github.com/karpathy/autor… Part code, part sci-fi, and a pinch of psychosis :)
English