[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

[@deredleritt3r](/creator/twitter/deredleritt3r)
"Sholto Douglas (Anthropic): "Over the last year RL has finally allowed us to take a feedback loop and turn it into a model that is at least as good as the best humans at a given thing in a narrow domain. And you're seeing that with mathematics and competition code which are the two domains most amendable to this - where rapidly the models are becoming incredibly competent competition mathematicians and competition coders. There's nothing intrinsically different about competition code and math. It's just that they're really more amenable to RL than any other domain. But importantly they"  
[X Link](https://x.com/deredleritt3r/status/1973812141846143241) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-02T18:07Z 5481 followers, 129.2K engagements


"August 2025: Oxford and Cambridge mathematicians publish a paper entitled "No LLM Solved Yu Tsumura's 554th Problem". They gave this problem to o3 Pro Gemini XXX Deep Think Claude Opus X (Extended Thinking) and other models with instructions to "not perform a web search to solve the problem". No LLM could solve it. The paper smugly claims: "We show contrary to the optimism about LLMs problem-solving abilities fueled by the recent gold medals that were attained that a problem existsYu Tsumuras 554th problemthat a) is within the scope of an IMO problem in terms of proof sophistication b) is not"  
[X Link](https://x.com/deredleritt3r/status/1974862963442868228) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-05T15:43Z 5481 followers, 220K engagements


"Meta is close to sealing a $30B financing deal for its 5GW Hyperion data center in Louisiana. - Target data center completion date: 2029 - Meta will be developer operator and tenant of the project - Total financing is $27B in debt $2.5B in equity. Meta will form a special purpose vehicle (SPV) that will issue the debt. Meta will own XX% of the SPV; the remainder will be owned by Blue Owl Capital - Debt matures in 2049; will be priced at XXX bps over U.S. Treasuries This is a similar financing structure to xAI's recent $20B deal for its Colossus data center"  
[X Link](https://x.com/deredleritt3r/status/1978986779987341528) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-17T00:49Z 5481 followers, 45.8K engagements


"@sama Initial feedback: Need this on Windows and Android yesterday. Also Sora X on Android. 🙏"  
[X Link](https://x.com/deredleritt3r/status/1980691769433157691) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-21T17:44Z 5481 followers, 1690 engagements


"@herbiebradley This structure keeps almost the entire deal off Meta's balance sheet. The debt isn't Meta's so it doesn't affect Meta's credit rating and otherwise has no impact on Meta's borrowing capabilities. I'm sure there are other advantages too - maybe tax etc"  
[X Link](https://x.com/deredleritt3r/status/1979010568758378709) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-17T02:24Z 5454 followers, XXX engagements


"For those who still don't get it: - Researchers working at the frontier labs are genuinely convinced that transformative AI is coming. - They spend their time worrying about the social/political consequences of transformative AI not the "AI bubble". - Recursive self-improvement is on the table and is the subject of serious discussions"  
[X Link](https://x.com/deredleritt3r/status/1976699624481083508) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-10T17:21Z 5477 followers, 346.2K engagements


"More details on Meta's new data center deal with Blue Owl for those interested: - Blue Owl Capital has created the SPV. The SPV will own an XX% interest in a joint venture (JV) with Meta; Meta will own the remaining 20%. - The JV will own the landlord entity. The tenant will be owned by Meta. - The data center owned by the landlord will initially (by June X 2029) be 2.064GW. As noted in my previous post it has been reported that there are plans to eventually expand this to 5GW. - Meta will contribute $5.76B to the JV. Blue Owl will contribute $23.03B. Total committed to build the data center"  
[X Link](https://x.com/deredleritt3r/status/1979368465849840029) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-18T02:06Z 5482 followers, 5387 engagements


"Now that everyone on X is bearish on achieving AGI soon because current models lack continual learning it's time to repost these: Dario Amodei (July 2025): "One thing we learned in AI is whenever it feels like there's some fundamental obstacle - like two years ago we thought there was this fundamental obstacle around reasoning and it turned out just to just be RL. Without being too specific I think and we already have maybe some evidence to suggest that continual learning is another of those problems that is not as difficult as it seems that will fall to scale plus a slightly different way of"  
[X Link](https://x.com/deredleritt3r/status/1979534425571402124) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-18T13:05Z 5482 followers, 67.3K engagements


"All those doubts will be washed away by scale like tears in the rain"  
[X Link](https://x.com/deredleritt3r/status/1979691671941226770) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-18T23:30Z 5481 followers, 7157 engagements


"Bloomberg: OpenAI has contracted with XXX ex-investment bankers to provide data for training an AI system codenamed "Project Mercury" that builds financial models and "aims to replace entry-level tasks at investment banks". Participants are asked to build Excel financial models for projects like restructurings and IPOs. They are also granted early access to the model. The interview process for this job apparently involves no human interaction - a 20-minute interview with an AI chatbot and two tests (financial statements and modeling) are all that's required"  
[X Link](https://x.com/deredleritt3r/status/1980642740728684822) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-21T14:30Z 5481 followers, 12.6K engagements


"Anthropic is in discussions with Google for a deal to provide Anthropic with cloud computing power valued in the "high tens billions of dollars". As a reminder Google is an existing investor in Anthropic ($2B invested in 2023 and an additional $1B invested this year)"  
[X Link](https://x.com/deredleritt3r/status/1980755155155886503) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-21T21:56Z 5481 followers, 1145 engagements


"@OpenAI The only reason to access ChatGPT via WhatsApp was for airplane flights that offer free WhatsApp messaging. Sad that this use case is going away"  
[X Link](https://x.com/deredleritt3r/status/1980795689581048255) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-22T00:37Z 5481 followers, 24.8K engagements


"Greg Brockman on Move 37: "I do think next year we will have some incredible models. The milestone I am most excited about is having models that can solve hard problems. And the the analogy I like to make is think back to AlphaGo you know Move XX that changes people's understanding of the game. Imagine that in coding imagine that in material science imagine that in medicine. Having real breakthroughs that are either potentially the AI by itself or the AI assisted with top humans. I think we'll start seeing that.""  
[X Link](https://x.com/deredleritt3r/status/1976056338342871327) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-08T22:45Z 5459 followers, 78.4K engagements


"Greg Brockman on using OpenAI models to optimize the chips OpenAI is developing with Broadcom: "We've been able to apply our own models to designing this chip. We've been able to get massive area reductions. You take components that humans have already optimized and just pour compute into it and the model comes up with its own optimizations. We're at the point now where I don't think any of the optimizations we have are ones that human designers couldn't have come up with. Usually our experts take a look at it later and say yeah this was on my list but it was a list of XX things that would"  
[X Link](https://x.com/deredleritt3r/status/1977877469026795629) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-13T23:21Z 5459 followers, 42.5K engagements


"@daniel_mac8 Great post. There is no amount of money one should not be rationally willing to expend in order to achieve post-scarcity"  
[X Link](https://x.com/deredleritt3r/status/1980056144329814104) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-19T23:39Z 5460 followers, XXX engagements


"AI is the superhighway to post-scarcity. The ultimate goal is energy and intelligence too cheap to meter robots building other robots at essentially no cost + a drastically increased rate of scientific discovery leading to abundance of most goods and services. If one accepts post-scarcity as a realistic possibility (which many researchers working at the frontier labs do; e.g. Sam Altman thinks it will be achieved in the 2030s) one should then conclude that the industry's investment in compute is very likely not a "bubble". There is no amount of capital that rational economic actors should not"  
[X Link](https://x.com/deredleritt3r/status/1977497131339653441) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-12T22:10Z 5466 followers, 59.1K engagements


"We are indeed at a crossroads with LLMs. Where we are: GPT-5 Pro is able to generate minor novel advancements in math (but probably nothing a human expert wouldn't be able to generate). OpenAI's models can propose optimizations for its chips but only of the kind human experts already had on their to-do list. Agentic coding tools work for many use cases but e.g. Andrej Karpathy found them unhelpful in creating nanochat "possibly because the repo is too far off the data distribution". Where we could go (optimistic scenario): Better models finally enable Move XX across many scientific"  
[X Link](https://x.com/deredleritt3r/status/1978159312863105041) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-14T18:01Z 5466 followers, 97.4K engagements


"Google and Yale scientists have trained an LLM that has generated a novel hypothesis about cancer cellular behavior. This prediction was confirmed multiple times in vitro. - "What made this prediction so exciting was that it was a novel idea. Although CK2 has been implicated in many cellular functions including as a modulator of the immune system inhibiting CK2 via silmitasertib has not been reported in the literature to explicitly enhance MHC-I expression or antigen presentation. This highlights that the model was generating a new testable hypothesis and not just repeating known facts." The"  
[X Link](https://x.com/deredleritt3r/status/1978541659869597978) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-15T19:21Z 5481 followers, 966K engagements


"Just to recap: We found out today that an LLM that fits on a high-end consumer GPU when trained on specific biological data can discover a novel method to make cancer tumors more responsive to immunotherapy. Confirmed novel discovery (not present in existing literature). Experimentally validated in living cells. This is AI generating novel science. The moment has finally arrived"  
[X Link](https://x.com/deredleritt3r/status/1978561622932164905) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-15T20:40Z 5481 followers, 930K engagements


"@JusBili They tested the hypothesis in the lab:"  
[X Link](https://x.com/deredleritt3r/status/1978580976067387571) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-15T21:57Z 5480 followers, 14K engagements


"GPT-5 has been used to find solutions in existing literature to XX Erdos problems that were previously considered unsolved. Terence Tao commenting on this: "I am increasingly of the opinion that the most productive near-term adoptions of AI in mathematics will primarily come. from using medium-powered tools to accelerate and scale up more mundane and time-consuming but still essential research tasks using the accumulated human experience. In such use cases the output of the AI tool could also have been produced (with increased expenditure of time and attention) by a human expert - but this is"  
[X Link](https://x.com/deredleritt3r/status/1979250915610890750) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-17T18:19Z 5481 followers, 171.6K engagements


"@firstadopter I think it just means that Anthropic needs to diversify its sources of compute - the same thing that e.g. OpenAI is doing right now. But Dario can't go to Jensen because he's pissed Jensen off. So the only other place he has left is Google - a competitor()"  
[X Link](https://x.com/deredleritt3r/status/1980756663066841522) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-21T22:02Z 5481 followers, 1060 engagements


"@sdmat123 @firstadopter Correct. But still imagine if we do wind up in a race for "the lightcone" and it's Google vs. Anthropic at the finish line"  
[X Link](https://x.com/deredleritt3r/status/1980760306352599509) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-21T22:17Z 5481 followers, XXX engagements


"@emollick The biggest open question in AI for me right now is whether more powerful future LLMs can in fact eventually produce the "Move 37" in scientific discovery. The frontier labs all seem to be convinced that the answer is "Yes". But let's wait and see"  
[X Link](https://x.com/deredleritt3r/status/1980802588896567709) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-22T01:05Z 5481 followers, XXX engagements


"Here is yet another case of an expert mathematician using GPT-5 Pro as a tool in a lengthy iterative process of solving an open problem (this time in convex optimization). GPT-5 Pro did not come up with a solution from a single prompt. The arguments it generated were wrong XX% of the time. Yet it "significantly accelerated" finding the solution by helping to eliminate dead-end approaches. And ultimately "the key successful steps were suggested by ChatGPT". Professor Ryu: "In my view this result is already publishable in a respectable optimization theory journal.""  
[X Link](https://x.com/deredleritt3r/status/1980800267818782872) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-22T00:55Z 5481 followers, 60.6K engagements


"@solarapparition I think I'll wind up being a very late adopter of agentic browser tools for the reasons you listed. I just instinctively don't trust them to be secure at this early stage"  
[X Link](https://x.com/deredleritt3r/status/1980809140563026252) [@deredleritt3r](/creator/x/deredleritt3r) 2025-10-22T01:31Z 5481 followers, XXX engagements