# ![@deredleritt3r Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::1744055410108080128.png) @deredleritt3r prinz

prinz posts on X about ai, open ai, anthropic, in the the most. They currently have [------] followers and [---] posts still getting attention that total [-------] engagements in the last [--] hours.

### Engagements: [-------] [#](/creator/twitter::1744055410108080128/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1744055410108080128/c:line/m:interactions.svg)

- [--] Week [---------] +299%
- [--] Month [---------] +424%
- [--] Months [----------] +826%
- [--] Year [----------] +3,802%

### Mentions: [--] [#](/creator/twitter::1744055410108080128/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1744055410108080128/c:line/m:posts_active.svg)

- [--] Month [--] -57%
- [--] Months [---] +111%
- [--] Year [---] -8.30%

### Followers: [------] [#](/creator/twitter::1744055410108080128/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1744055410108080128/c:line/m:followers.svg)

- [--] Week [------] +4.20%
- [--] Month [------] +11%
- [--] Months [------] +308%
- [--] Year [------] +1,038%

### CreatorRank: [------] [#](/creator/twitter::1744055410108080128/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::1744055410108080128/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[technology brands](/list/technology-brands)  33.65% [finance](/list/finance)  7.69% [stocks](/list/stocks)  5.77% [social networks](/list/social-networks)  0.96%

**Social topic influence**
[ai](/topic/ai) #2311, [open ai](/topic/open-ai) #157, [anthropic](/topic/anthropic) #2405, [in the](/topic/in-the) 9.62%, [model](/topic/model) #924, [agi](/topic/agi) #25, [search](/topic/search) 3.85%, [the first](/topic/the-first) 3.85%, [ft](/topic/ft) #5174, [$googl](/topic/$googl) 3.85%

**Top accounts mentioned or mentioned by**
[@garymarcus](/creator/undefined) [@emollick](/creator/undefined) [@wrwagox](/creator/undefined) [@noeflandre](/creator/undefined) [@vigxyz](/creator/undefined) [@davidkasten](/creator/undefined) [@danielmac8](/creator/undefined) [@ft](/creator/undefined) [@mreismaus](/creator/undefined) [@lefthanddraft](/creator/undefined) [@lower_votingage](/creator/undefined) [@xmaneyman](/creator/undefined) [@precedent_vice](/creator/undefined) [@_simonsmith](/creator/undefined) [@fangyi11101](/creator/undefined) [@bunagayafrost](/creator/undefined) [@iz9smi0cmu90ihn](/creator/undefined) [@snowflakeflamer](/creator/undefined) [@databaseddad](/creator/undefined) [@bobasoynpc](/creator/undefined)

**Top assets mentioned**
[Alphabet Inc Class A (GOOGL)](/topic/$googl) [Microsoft Corp. (MSFT)](/topic/microsoft)
### Top Social Posts
Top posts by engagements in the last [--] hours

"And now for something completely different. Introducing: prinzbench. Over the past few months I have been increasingly using LLMs for legal work primarily legal research and "needle-in-the-haystack" search queries. This is economically valuable work of the kind performed by many lawyers on a daily basis. It is also quite distinct from the benchmarks that the frontier labs typically "hill-climb" - i.e. math and coding. And importantly correct legal analysis is not easily verifiable. There is no code to compile there is no mathematical proof that either works or doesn't. The only way to know"  
[X Link](https://x.com/deredleritt3r/status/2013979845378580684)  2026-01-21T14:19Z [----] followers, 54.3K engagements


"This is what the first stage of takeoff looks like. End-to-end AI research (other than generation of actual research ideas) is next. @chatgpt21 Can confirm - we made the OpenAI MCP server in [--] days (and scale proof) theres Sora android app in [--] weeks and a lot more internal tooling built by codex reviewed by codex. Hard to see how OpenAI could still ship at the speed it does without codex. Anecdotally I spend a @chatgpt21 Can confirm - we made the OpenAI MCP server in [--] days (and scale proof) theres Sora android app in [--] weeks and a lot more internal tooling built by codex reviewed by codex."  
[X Link](https://x.com/deredleritt3r/status/2015408902976766064)  2026-01-25T12:58Z [----] followers, 95K engagements


"I've been looking at coverage of Dario Amodei's remarks at Davos. Lots of articles about his views on export controls impact of AI on employment etc. Basically ZERO coverage of when he said: "We're building AI that can research how to build even smarter AI; trying to close the loop; it will go faster than people imagine; expect rapid progress this year." Even the timeline here on X doesn't seem to fully get it. Automation of AI research is *the* most important thing happening in the field of AI right now and by some distance. https://twitter.com/i/web/status/2015417344194150558"  
[X Link](https://x.com/deredleritt3r/status/2015417344194150558)  2026-01-25T13:31Z [----] followers, 58.9K engagements


"Added to prinzbench: Qwen Q3-Max and Kimi K2.5-Thinking. Kimi K2.5-Thinking performs as well as the Gemini [--] models and is tied with Gemini [--] Pro for third place. Qwen Q3-Max is not nearly as impressive"  
[X Link](https://x.com/deredleritt3r/status/2016867270929109365)  2026-01-29T13:33Z [----] followers, 25.8K engagements


"The real fun starts when interactions among AI agents are no longer pure wordcelism but instead guided by market forces. Let them cooperate to create economic value. Let them compete. May the best .md files win"  
[X Link](https://x.com/deredleritt3r/status/2017441145341550972)  2026-01-31T03:33Z [----] followers, [----] engagements


"January [--] 2025: near the singularity; unclear which side June [--] 2025: We are past the event horizon; the takeoff has started. Somewhere in-between "it" happened"  
[X Link](https://x.com/deredleritt3r/status/2018212657153650792)  2026-02-02T06:39Z [----] followers, 10.7K engagements


"Scheduled tasks in Codex. This is huge. For example I should now easily be able to ask Codex to run a daily search for newly introduced legislation. Or new publicly available agreements of a certain type. Or new guidance issued by regulators. So easy even a lawyer can do it. OpenAI announces new automation feature in the new Codex App. This is going to be huge https://t.co/zjLOLJB2yl OpenAI announces new automation feature in the new Codex App. This is going to be huge https://t.co/zjLOLJB2yl"  
[X Link](https://x.com/deredleritt3r/status/2018408770808807785)  2026-02-02T19:38Z [----] followers, 56.9K engagements


""OpenAI chief research officer Mark Chen tells Forbes that in the year ahead it hopes to develop an AI researcher 'intern' that can help his team accelerate its ideas. 'We are heading toward a system that will be capable of doing innovation on its own' Altman says. 'I dont think most of the world has internalized what thats going to mean.'" As CEO of OpenAI the 40-year-old billionaire helped unleash ChatGPT pushing artificial intelligence into the mainstream and reshaping the global economy. Now a new father with another baby on the way hes building the world his kids will one day inherit."  
[X Link](https://x.com/deredleritt3r/status/2018760613481119784)  2026-02-03T18:56Z [----] followers, 37.5K engagements


""Were in the business of creating an automated scientist" 'Nuff said. How does OpenAI balance long-term research bets with product-forward research fundamentals Ive been getting this question a lot lately usually framed as a suggestion that Jakub (@merettm) and I are pushing an increasingly product-focused agenda. That characterization is How does OpenAI balance long-term research bets with product-forward research fundamentals Ive been getting this question a lot lately usually framed as a suggestion that Jakub (@merettm) and I are pushing an increasingly product-focused agenda. That"  
[X Link](https://x.com/deredleritt3r/status/2018780973186642018)  2026-02-03T20:17Z [----] followers, 11.8K engagements


"@vig_xyz I'm a lawyer. I have unlimited access to GPT-5.2 Pro at work. My understanding is that the vast majority of lawyers at other law firms don't. This is completely wild to me but is the reality right now"  
[X Link](https://x.com/deredleritt3r/status/2019166171959054559)  2026-02-04T21:48Z [----] followers, 10.5K engagements


"@FangYi11101 @vig_xyz Nope. I've heard from a few people who use Harvey that it's not so great - but can't tell whether this is the fault of the user (not knowing how to properly use AI) the law firm (not properly implementing Harvey) or Harvey itself"  
[X Link](https://x.com/deredleritt3r/status/2019180206435492332)  2026-02-04T22:44Z [----] followers, [---] engagements


"We have some yes. It's the more junior people who use these tools TBH - I've not touched them. Fir diligence I'd personally prefer GPT-5.2 Pro (but may need a harness to have it run longer). I've used it to get regulatory analysis for some diligence documents to good effect. https://twitter.com/i/web/status/2019181187957219573 https://twitter.com/i/web/status/2019181187957219573"  
[X Link](https://x.com/deredleritt3r/status/2019181187957219573)  2026-02-04T22:48Z [----] followers, [--] engagements


"@_simonsmith @vig_xyz Really interesting and I wonder whether this is happening at scale in other industries as well (the smaller players jumping directly to ChatGPT Enterprise)"  
[X Link](https://x.com/deredleritt3r/status/2019186215535198518)  2026-02-04T23:08Z [----] followers, [---] engagements


"@voooooogel You are thinking about this in exactly the right (rational) way but most people are anything but rational and instead think "woaw this looks coooool ❤" and click the big red "Purchase" button"  
[X Link](https://x.com/deredleritt3r/status/2019218674444640285)  2026-02-05T01:17Z [----] followers, [---] engagements


"@wrwagox @vig_xyz We have a ChatGPT Enterprise subscription at work. OpenAI doesn't see any client data and doesn't train on our prompts. I would *never* upload confidential client information to any other AI (like my personal ChatGPT account)"  
[X Link](https://x.com/deredleritt3r/status/2019250046555095168)  2026-02-05T03:21Z [----] followers, [--] engagements


"@MrEismaus Thanksgiving and Christmas/New Year break ruined the tempo. Otherwise I imagine we would have had [---] Codex last month"  
[X Link](https://x.com/deredleritt3r/status/2019491604235710955)  2026-02-05T19:21Z [----] followers, [---] engagements


"@_sholtodouglas Is a "software-only singularity" the final outcome of "powerful AI" In other words do you think that powerful AI arrives in around [--] months (Anthropic's prior prediction) and leads to the software-only singularity in around another 1-2 years (2028-2029)"  
[X Link](https://x.com/deredleritt3r/status/2019588680835551328)  2026-02-06T01:47Z [----] followers, [----] engagements


"@mattyglesias This is just the first step. Automation of AI research and RSI are on the horizon. Too few paid attention to Dario Amodei's remarks at Davos: https://x.com/deredleritt3r/status/2013613671704924640s=20 Dario Amodei on reaching "a model that can do everything a human can do at a level of a Novel laureate across many fields" aka AGI: "I don't think it's far off. The mechanism by which I imagined it would happen is that we would have models that are good at coding and AI https://x.com/deredleritt3r/status/2013613671704924640s=20 Dario Amodei on reaching "a model that can do"  
[X Link](https://x.com/deredleritt3r/status/2019898725431415067)  2026-02-06T22:19Z [----] followers, [----] engagements


"@Bunagayafrost Anthropic is not nearly that simple It's not a c-corp but a public benefit corporation. And there's also an entity in the structure called the Long-Term Benefit Trust that makes things significantly more complicated: https://www.anthropic.com/news/the-long-term-benefit-trust https://www.anthropic.com/news/the-long-term-benefit-trust"  
[X Link](https://x.com/deredleritt3r/status/2019942036703412734)  2026-02-07T01:11Z [----] followers, [---] engagements


"@apples_jimmy Agreed. It's a hard question. Seems to be something for the two labs to consider internally"  
[X Link](https://x.com/deredleritt3r/status/2019942298075668556)  2026-02-07T01:12Z [----] followers, [---] engagements


"Around a month or two before OpenAI released GPT-5 we got news that Anthropic was rapidly growing its enterprise revenue. Claude Code Opus [--] and Sonnet [--] were a jet engine. OpenAI could not afford to sit back and do nothing. OpenAI handled this masterfully. GPT-5 was faster and cheaper than o3 but just as smart. It had a colder "let's get to work" personality (which was despised by the GPT-4o addicts). And most importantly it nearly eliminated hallucinations - which plagued o3 and made it very hard to use at work. Just a month later a new Codex model based on GPT-5 took the world by storm."  
[X Link](https://x.com/deredleritt3r/status/2020268218057719955)  2026-02-07T22:47Z [----] followers, [----] engagements


"@eigenrobot I can see this happening for jobs involving physical labor assuming that we get the "software-only singularity" but robotics lags far behind"  
[X Link](https://x.com/deredleritt3r/status/2020307586336342143)  2026-02-08T01:23Z [----] followers, [---] engagements


"@iz9Smi0cmU90iHN With GPT-5.3 seemingly close to release I think I might just wait to test GPT-5.3 Pro instead"  
[X Link](https://x.com/deredleritt3r/status/2020549064031367321)  2026-02-08T17:23Z [----] followers, [---] engagements


"Broadly speaking Opus [---] is very similar to Anthropic's other models. In the COT I occasionally see it zero in on the correct source where the relevant info is found but fail to open the source of fetch the info (this seems to be a problem with tool use and/or failure to allocate sufficient test-time compute to the task). But it's more than that. All too often the model just isn't sufficiently comprehensive in examining the relevant sources of information and winds up missing a key statutory reference in its reasoning. And when it does locate the relevant sources of information its reasoning"  
[X Link](https://x.com/deredleritt3r/status/2020576613457363206)  2026-02-08T19:12Z [----] followers, [----] engagements


"@snowflakeflamer @David_Kasten I haven't been able to directly compare them but can attest that GPT-5.2 Pro is great for legal research and analysis"  
[X Link](https://x.com/deredleritt3r/status/2020914959027196024)  2026-02-09T17:37Z [----] followers, [--] engagements


"@mauddweeb GPT-4o wasn't a frontier model the moment o1-preview was released (September [----] [--] months ago)"  
[X Link](https://x.com/deredleritt3r/status/2020946435093803326)  2026-02-09T19:42Z [----] followers, [--] engagements


"@NicolasSerna314 Probably nothing. It's very common for people to switch jobs for various reasons (especially early in the year) and it's always been a myth that Anthropic is somehow exempt from this"  
[X Link](https://x.com/deredleritt3r/status/2021017809635963352)  2026-02-10T00:26Z [----] followers, [----] engagements


"@daniel_mac8 Yes sounds like it"  
[X Link](https://x.com/deredleritt3r/status/2021027664782610910)  2026-02-10T01:05Z [----] followers, [----] engagements


"So easy for a client to upload relevant company info past coverage reports from insurance brokers claims history and just about anything else even remotely relevant into context have the data analyzed by GPT-5.2 Thinking (or even Pro) feed the resulting "clean" information to multiple insurance broker apps within ChatGPT and then have the model return the best quotes along with analysis comparing them all in one output. Turns highly customized client-specific business into a volume business instantly. Insurance brokers that aren't prepared to compete in this fashion by offering the best deals"  
[X Link](https://x.com/deredleritt3r/status/2021052479828267370)  2026-02-10T02:43Z [----] followers, [----] engagements


"This doesn't surprise me in the slightest and will happen more and more especially as: (i) models continue improving; (ii) clients begin figuring out which models to use and which to avoid (hint: use GPT-5.2 Pro); and (iii) the frontier labs build out agentic tools that specifically help lawyers in the workplace. https://twitter.com/i/web/status/2021071484064571784 https://twitter.com/i/web/status/2021071484064571784"  
[X Link](https://x.com/deredleritt3r/status/2021071484064571784)  2026-02-10T03:59Z [----] followers, [---] engagements


"I sympathize but it's actually much easier to get a low-limit credit card specifically for your agents than to pay in crypto. Remember: credit card charges can be disputed. If your agent gets defrauded into making a payment that payment can be fought and reversed. With crypto once your stablecoins are gone they're gone forever. Not to mention that each agent would have to use a hot wallet (can't use a hardware wallet like a Ledger if you have no hands). Hot wallets are notoriously insecure and prone to hacking. https://twitter.com/i/web/status/2021074280813936941"  
[X Link](https://x.com/deredleritt3r/status/2021074280813936941)  2026-02-10T04:10Z [----] followers, [----] engagements


"I'd say that fraud is quite limited (same as it is in real life). For example (and I'm sure there are others) if my agent spends money on X but never receives X (or receives X that is materially different in quality from whatever was promised) that's fraud. Prompt injections are also obviously fraud. The agent acts under authority from me to perform certain actions. That authority needs to be sufficiently tight so as to ensure that the agent doesn't waste money on unnecessary things. If I've failed to set the guidelines appropriately that should be my problem. (And yes I expect that this will"  
[X Link](https://x.com/deredleritt3r/status/2021080682638750014)  2026-02-10T04:35Z [----] followers, [---] engagements


"@SebastienBubeck Yes. Accelerando"  
[X Link](https://x.com/deredleritt3r/status/2021081618958442781)  2026-02-10T04:39Z [----] followers, [----] engagements


"@DataBasedDad @David_Kasten Yes. We have ChatGPT Enterprise at work. OpenAI does not see our prompts or train its models on our data"  
[X Link](https://x.com/deredleritt3r/status/2021243284216012999)  2026-02-10T15:22Z [----] followers, [--] engagements


"@NoeFlandre @David_Kasten This post should give you a fairly good idea: https://x.com/deredleritt3r/status/2002064109223752163 @MrEismaus @lemire My answer will have to be nuanced: [--]. For legal research and analysis GPT-5.x Pro is stellar and GPT-5.x Thinking is very good. All other models (including Opus [---] Gemini [--] Pro Gemini [--] Flash Grok) are unusable. Even for the GPT-5.x models there are quite a few caveats. https://x.com/deredleritt3r/status/2002064109223752163 @MrEismaus @lemire My answer will have to be nuanced: [--]. For legal research and analysis GPT-5.x Pro is stellar and GPT-5.x"  
[X Link](https://x.com/deredleritt3r/status/2021274006142276079)  2026-02-10T17:24Z [----] followers, [--] engagements


"Thank you I saw that and it looks interesting. Unfortunately nothing is okay to use in legal if my firm hasn't specifically approved it. I would never upload client data anywhere other than a model in my ChatGPT Enterptise plan or (in theory) a local model running on my own GPU. https://twitter.com/i/web/status/2021279109523882025 https://twitter.com/i/web/status/2021279109523882025"  
[X Link](https://x.com/deredleritt3r/status/2021279109523882025)  2026-02-10T17:44Z [----] followers, [--] engagements


"I am very thankful that we have two excellent frontier labs OpenAI and Anthropic racing towards the same goal for the betterment of humanity. And like Jerry Tworek I would be *much* more worried about the potential negative implications of automation of AI research and RSI if only *one* frontier lab were pursuing it. All that said attacking one another over an inconsequential subject like ads seems pointless and unnecessary. I personally would instead prefer that each lab's progress towards AGI and publicly released models do the talking. As I see it right now future dystopia is a single"  
[X Link](https://x.com/anyuser/status/2019157582494199819)  2026-02-04T21:14Z 10.2K followers, [----] engagements


"@ramez Breaking down. Sad to see. BTW his own team was using Claude Code up until a few weeks ago"  
[X Link](https://x.com/deredleritt3r/status/2022098258483691726)  2026-02-12T23:59Z 10.2K followers, [----] engagements


""It seems almost obvious that using more test time compute will make it easier to solve problems but this simple concept has surprising implications for market competition capital allocation contests between nations and scientific progress." I've written about the new role in The Prompt. Also releasing today is a paper my team worked on about the impacts of AI on international security; later today I'll follow up with a longer thread on the key ideas. https://t.co/mdgI1MbvST I've written about the new role in The Prompt. Also releasing today is a paper my team worked on about the impacts of"  
[X Link](https://x.com/deredleritt3r/status/2021652884459528403)  2026-02-11T18:29Z 10.2K followers, [----] engagements


"As a lawyer who uses LLMs every day at work I feel qualified to respond. First hallucinations are no longer a problem. Consistent with the prediction you quoted from [----] GPT-5.x almost never hallucinates. And overall the percentage of inaccurate responses I get from GPT-5.2 Pro is lower than the percentage of inaccurate responses I would get from a competent junior associate (yes fully accounting for hallucinations). Second people wildly overestimate the difficulty of most tasks performed by lawyers. The vast majority of the things we do are not nearly as challenging intellectually as"  
[X Link](https://x.com/deredleritt3r/status/2022314279668519076)  2026-02-13T14:17Z 10.2K followers, 227.6K engagements


"@edward_the6 Is this GPT-5.2 Thinking or Auto"  
[X Link](https://x.com/deredleritt3r/status/2022451674615591346)  2026-02-13T23:23Z 10.2K followers, [----] engagements


"(To that end in order to thoroughly mog the competition once and for all and show the world which lab really is the best I encourage OpenAI to publicly release the model that won gold on the ICPC and Anthropic to give us guidance on when we should expect the public release of continual learning.) https://twitter.com/i/web/status/2019157933179736273 https://twitter.com/i/web/status/2019157933179736273"  
[X Link](https://x.com/anyuser/status/2019157933179736273)  2026-02-04T21:15Z 10.2K followers, [----] engagements


"Anthropic's revenue (annualized): - January 2025: $1B - May: $3B - June: $4B - August: $5B - October: $7B - December: $8B to $10B -February 2026: $14B Our run-rate revenue is $14 billion and has grown over 10x in each of the past [--] years. This growth has been driven by our position as the intelligence platform of choice for enterprises and developers. Read more: https://t.co/aMRyOkFFSg Our run-rate revenue is $14 billion and has grown over 10x in each of the past [--] years. This growth has been driven by our position as the intelligence platform of choice for enterprises and developers. Read"  
[X Link](https://x.com/deredleritt3r/status/2022048290544722265)  2026-02-12T20:40Z 10.2K followers, 29.4K engagements


"Great point; let me explain this also. First it is correct that even with the very best model (GPT-5.2 Pro) I would *never* copy/paste its output into a legal brief. *Everything* generated by AI needs to be double-checked by an experienced lawyer - mostly not because of hallucinations (which as noted above are very rare for GPT-5.2 Pro but do occur very occasionally) but rather because the model may have reasoned poorly or not found some important precedent. But LLMs are not used in a vacuum. Today they are used as replacement *for a competent junior associate*. So yes - I have to check"  
[X Link](https://x.com/deredleritt3r/status/2022330823471505781)  2026-02-13T15:23Z 10.2K followers, 15.7K engagements


"Agreed on all counts. FWIW I think the holding in this particular case was correct. AI is not a licensed attorney and conversations with AI therefore should not be considered legally privileged. But I do think that the courts should consider enshrining a new privilege framework around conversations with AI"  
[X Link](https://x.com/deredleritt3r/status/2022403586785972438)  2026-02-13T20:12Z 10.2K followers, [---] engagements


"RT @polynoamial: There have been fair questions on whether LLM contributions to STEM are overhyped but I've spoken with physicists about t"  
[X Link](https://x.com/deredleritt3r/status/2022472730709242238)  2026-02-14T00:47Z 10.2K followers, [---] engagements


"@BobasoyNPC @GaryMarcus @FT @DamienCharlotin I'm not a litigator so can't really answer. I am fairly cautious with LLMs generally so probably wouldn't want the model to do work in real time when there's anything important on the line"  
[X Link](https://x.com/deredleritt3r/status/2022485801846870066)  2026-02-14T01:39Z 10.2K followers, [---] engagements


"Indeed. I am using GPT-5.2 Pro at work every day and find it to be very productivity-enhancing. Many other senior lawyers at my firm have never touched our ChatGPT Enterprise plan or at most have used the Auto mode to edit an e-mail or two. Meanwhile my assumption is that most lawyers at other firms don't have access to anything even remotely as good as GPT-5.2 Pro. It's a strange time"  
[X Link](https://x.com/deredleritt3r/status/2020901515821469813)  2026-02-09T16:44Z [----] followers, [----] engagements


"@NoeFlandre @David_Kasten Legal documents live in MS Word (just the way it is) so I've been waiting for an agentic interface to arrive to "vibe-code" agreements in Word. We don't have access to Codex or Prism at work so I don't have too much latitude to experiment until that happens"  
[X Link](https://x.com/deredleritt3r/status/2021277424701304924)  2026-02-10T17:37Z [----] followers, [--] engagements


"@lefthanddraft High number of people chatting with it (enough to form a community around chatting to this specific model) + it's past some (potentially even fairly low) sycophancy threshold"  
[X Link](https://x.com/deredleritt3r/status/2021650493249135009)  2026-02-11T18:20Z 10.2K followers, [----] engagements


"@MillionInt I find it crazy that the race to AGI is happening *literally during my lifetime*; what are the chances of that even happening Makes me wonder somewhat whether this is a simulation and how much of this is preordained (including plot armor for me personally)"  
[X Link](https://x.com/deredleritt3r/status/2021711868033704330)  2026-02-11T22:24Z 10.2K followers, [----] engagements


"Yes good point. I actually personally feel that being alive and aware during the takeoff is more interesting somehow than being born [----] years post-singularity. But there is nothing rational about this and you're right that one could make an argument that it's living and dying pre-take-off that is incredibly statistically unlikely. https://twitter.com/i/web/status/2021758401546424709 https://twitter.com/i/web/status/2021758401546424709"  
[X Link](https://x.com/deredleritt3r/status/2021758401546424709)  2026-02-12T01:29Z 10.2K followers, [--] engagements


"@emollick What do you recommend as the few notable exceptions"  
[X Link](https://x.com/deredleritt3r/status/2019217556221943888)  2026-02-05T01:12Z 10.2K followers, [----] engagements


"My view re: claims that "AI will be able to replace most tasks performed by lawyers in 12-18 months": @GaryMarcus @FT As a lawyer who uses LLMs every day at work I feel qualified to respond. First hallucinations are no longer a problem. Consistent with the prediction you quoted from [----] GPT-5.x almost never hallucinates. And overall the percentage of inaccurate responses I get from GPT-5.2 @GaryMarcus @FT As a lawyer who uses LLMs every day at work I feel qualified to respond. First hallucinations are no longer a problem. Consistent with the prediction you quoted from [----] GPT-5.x almost"  
[X Link](https://x.com/deredleritt3r/status/2022322226268131343)  2026-02-13T14:49Z 10.2K followers, 24.5K engagements


"@adonis_singh Sebastian Bubeck recently said that OpenAI had used a fairly simple scaffold to have GPT-5 run for two days to solve one of the math problems in the November "Early science acceleration experiments with GPT-5" paper. https://x.com/deredleritt3r/status/2018370801783755218 "The scaffold is not that complicated. You can imagine building some multi-agent system that is making API calls to GPT-5 where there's one agent that proposes a proof another agent that kind of critiques the proof and so on and so forth." https://x.com/deredleritt3r/status/2018370801783755218 "The scaffold is"  
[X Link](https://x.com/deredleritt3r/status/2022488861817885005)  2026-02-14T01:51Z 10.2K followers, [----] engagements


"@MackinacCap @GaryMarcus @FT Yes totally possible. In practice don't even need to do that since OpenAI offers a ChatGPT Enterprise plan"  
[X Link](https://x.com/deredleritt3r/status/2022321746716643719)  2026-02-13T14:47Z 10.2K followers, [----] engagements


"Sam Altman: - "If you're a sophomore you'll graduate to a world with AGI in it." - The next ChatGPT moment is "the same level of ability to get stuff done as Codex is for coding for all knowledge work which I don't think is very far away." This implies that AGI will be achieved by May [----] which is also close to OpenAI's timeline for achieving end-to-end automated AI research (by March 2028). Am judging @hackwithtrees Stanford samas takes on: - the literal first day of OpenAI - if youre a sophomore youll graduate to a world with AGI - why @MattPRDs Moltbook matters - what you should work on"  
[X Link](https://x.com/deredleritt3r/status/2022678584553255200)  2026-02-14T14:25Z 10.2K followers, 49.4K engagements


"@hamandcheese These IPOs need to happen soon. Dario recently wrote about his fear that much of the wealth generated by AI will remain trapped in the Silicon Valley. The public markets fix this"  
[X Link](https://x.com/deredleritt3r/status/2019459021774663940)  2026-02-05T17:12Z 10.2K followers, [----] engagements


"@Tril1boswagginz I wasn't going to test Codex [---] since it's a specialized coding/agentic model but I suppose I could do so if there's enough interest"  
[X Link](https://x.com/deredleritt3r/status/2020544294885736708)  2026-02-08T17:04Z 10.2K followers, [----] engagements


"@davidpiegaro Not yet. I'm a bit put off by the effort I would need to run [--] prompts through these models but might do it anyway. Also no results for Codex [---] just yet (another popular request)"  
[X Link](https://x.com/deredleritt3r/status/2022332550413504651)  2026-02-13T15:30Z 10.2K followers, [--] engagements


"METR (50% accuracy): GPT-5.1-Codex-Max = [--] hours [--] minutes This is [--] minutes longer than GPT-5"  
[X Link](https://x.com/deredleritt3r/status/1991245055017820236)  2025-11-19T20:39Z 10.2K followers, 779.5K engagements


"Opus [---] added to prinzbench. It scored 19/99 one point worse than Sonnet-4.5. The Anthropic models' utter inability to thoroughly search the web and use tools continues to destroy their performance on my benchmark"  
[X Link](https://x.com/deredleritt3r/status/2020525038110195886)  2026-02-08T15:48Z 10.2K followers, 32.8K engagements


"@willdepue @ericmitchellai @yanndubs This was by far the most important change for me between o3 and GPT-5. The GPT-5.x Pro models specifically have been absolutely superb"  
[X Link](https://x.com/deredleritt3r/status/2022338502348964083)  2026-02-13T15:54Z 10.2K followers, [---] engagements


"To clarify with GPT-5.2 Pro hallucinations these days are an *extremely* small concern because they *almost never* happen. It's far more likely that the model has failed to correctly analyze some statute or didn't take into account some potentially applicable case. In other words if I find myself watching the model like a hawk it's completely for the same reason as why I would find myself watching a junior associate like a hawk. The result of the above is that GPT-5.2 Pro is *extremely* useful in my legal work. And if the model is *extremely* useful in my legal work then I think it's fair to"  
[X Link](https://x.com/deredleritt3r/status/2022351028243317009)  2026-02-13T16:43Z 10.2K followers, [----] engagements


"Just a bit of nuance here: I think that GPT-5.2 Pro is good enough to replace a competent junior associate *on some tasks* (like legal research and analysis) which is very different from saying that it can replace *the entire job* of a junior associate (which I do not believe is the case). But yes - I think we will eventually expand the number of tasks and the levels of seniority to which this paradigm applies. https://twitter.com/i/web/status/2022355553469477199 https://twitter.com/i/web/status/2022355553469477199"  
[X Link](https://x.com/deredleritt3r/status/2022355553469477199)  2026-02-13T17:01Z 10.2K followers, [---] engagements


"It is absolutely correct that the hallucination rate is not 0% - but in my experience it's very very low. Also up until a few days ago Deep Research was based on the same model family as o3 which made it hallucinate significantly more than GPT-5.2 Thinking/Pro. This could have been the issue in your particular case. https://twitter.com/i/web/status/2022470367621591061 https://twitter.com/i/web/status/2022470367621591061"  
[X Link](https://x.com/deredleritt3r/status/2022470367621591061)  2026-02-14T00:38Z 10.2K followers, [---] engagements


"This position is responsible for implementing safeguards for specific AI capabilities when they reach High and Critical levels - including "AI self-improvement" I am extremely excited to welcome @dylanscandinaro to OpenAI as our Head of Preparedness. Things are about to move quite fast and we will be working with extremely powerful models soon. This will require commensurate safeguards to ensure we can continue to deliver tremendous I am extremely excited to welcome @dylanscandinaro to OpenAI as our Head of Preparedness. Things are about to move quite fast and we will be working with"  
[X Link](https://x.com/deredleritt3r/status/2018829951269834929)  2026-02-03T23:32Z 10.2K followers, [----] engagements


"We are probably building too few data centers As Rocks May Think: an interactive essay on thinking models automated research and where I think they are headed. Enjoy https://t.co/dcFYrXUQYg As Rocks May Think: an interactive essay on thinking models automated research and where I think they are headed. Enjoy https://t.co/dcFYrXUQYg"  
[X Link](https://x.com/anyuser/status/2019214267741008221)  2026-02-05T00:59Z 10.2K followers, 17.8K engagements


"Anthropic asked [--] of its researchers regarding the uplift they get from working with Opus [---]. - Reported uplift ranged from 30% to 700%(). - Mean uplift was 152%; median uplift was 100%. **Note that this survey was different from similar surveys performed for Sonnet [---] and Opus [---] in that the respondents were *not* Claude Code power users but rather a sample of Anthropic researchers and infra/maintenance roles"  
[X Link](https://x.com/anyuser/status/2019473920068579742)  2026-02-05T18:11Z 10.2K followers, 12.1K engagements


""GPT5.3Codex is our first model that was instrumental in creating itself. The Codex team used early versions to debug its own training manage its own deployment and diagnose test results and evaluations.""  
[X Link](https://x.com/deredleritt3r/status/2019475360438493597)  2026-02-05T18:17Z 10.2K followers, 38.8K engagements


"Cadence of recent Codex releases: - November 19: Codex [---] Max - December 18: Codex [---] - February 5: Codex [---] A significantly better model *every month*"  
[X Link](https://x.com/deredleritt3r/status/2019490579600511231)  2026-02-05T19:17Z 10.2K followers, [----] engagements


"@Noahpinion missing the point - software engineering is a special and a hard skill. it is the first barrier to recursive self improvement of artificial intelligence so it fell anyways. everything will follow in the order that they are bottlenecking recursive self improvement @Noahpinion missing the point - software engineering is a special and a hard skill. it is the first barrier to recursive self improvement of artificial intelligence so it fell anyways. everything will follow in the order that they are bottlenecking recursive self improvement"  
[X Link](https://x.com/deredleritt3r/status/2019802423691239502)  2026-02-06T15:56Z 10.2K followers, [----] engagements


"Microsoft is in an interesting position and is vastly underestimated. - can't really use OpenAI IP to develop AGI until OpenAI reaches AGI - will have access to AGI as soon as OpenAI declares it (but this right expires in 2030) As soon as OpenAI declares AGI (assuming this happens before 2030) Microsoft will have a copy of it and will need to further develop it separately from OpenAI. But who races faster once this happens If AI research is fully automated by then will this just depend on the available compute The biggest takeaway is this: when I write that OpenAI and Anthropic are racing"  
[X Link](https://x.com/deredleritt3r/status/2019930961417265523)  2026-02-07T00:27Z 10.2K followers, 25.8K engagements


""I believe the pace will surprise even many of those who are fully bought in and staying close to AI. . imagine models that are 10x faster smarter and more capable in specific domains. Then repeat the same exercise with the interfaces and products built on top of them." I spend quite a bit of time these days trying to imagine. Imagining AI progress continuing at its current velocity is already difficult. I must confess that imagining it relentlessly accelerating over the foreseeable future is almost beyond me. One thing Im missing from many takes on AIs impact is that were still so so early."  
[X Link](https://x.com/deredleritt3r/status/2020288544003043588)  2026-02-08T00:08Z 10.2K followers, 19.8K engagements


"https://github.com/prinz-ai/prinzbench https://github.com/prinz-ai/prinzbench"  
[X Link](https://x.com/deredleritt3r/status/2020525040681230611)  2026-02-08T15:48Z 10.2K followers, [----] engagements


"700 people just lost their jobs at the law firm Baker McKenzie based on "rethinking the way we work including through the use of AI". No lawyers impacted; cuts were made to "IT knowledge admin DEI leadership & learning secretarial marketing and design teams". Baker McKenzie just laid off [---] staff just under 10% because of Al. it's coming quick for our jobs. Baker McKenzie just laid off [---] staff just under 10% because of Al. it's coming quick for our jobs"  
[X Link](https://x.com/deredleritt3r/status/2021026582802928068)  2026-02-10T01:01Z 10.2K followers, 137K engagements


"Because the honest version sounds like I've lost my mind. And for a while I told myself that was a good enough reason to keep what's truly happening to myself. But the gap between what I've been saying and what is actually happening has gotten far too big. Very real. Exactly how I've been feeling for the past year or so and even more acutely over the past few months. https://twitter.com/i/web/status/2021298463359619358 https://twitter.com/i/web/status/2021298463359619358"  
[X Link](https://x.com/deredleritt3r/status/2021298463359619358)  2026-02-10T19:01Z 10.2K followers, 11.1K engagements


"It's rationally explainable - but also utterly insane on some level - that publicly declared pursuit of automation of AI research and recursive self-improvement by *two* frontier labs is subject only to half-hearted debate by the Twitter AI community and is not (yet) national news. I know we're all used to it now but it's so wild that recursive self improvement is actually happening now in some form and we're all just debating the pace This was a sci fi concept and some even questioned if it was possible at all I know we're all used to it now but it's so wild that recursive self improvement"  
[X Link](https://x.com/deredleritt3r/status/2021331650571882504)  2026-02-10T21:13Z 10.2K followers, 26.1K engagements


"@boazbaraktcs Agreed The hallucination rate even with GPT-5.2 Pro is definitely not 0% but it's very very low"  
[X Link](https://x.com/deredleritt3r/status/2022346175698063626)  2026-02-13T16:24Z 10.2K followers, [----] engagements


"@tobiaschneider This doesn't surprise me in the slightest. I would probably have this kind of work done by GPT-5.2 Pro and spot-checked by a lawyer (instead of having that lawyer do 100% of the work himself/herself)"  
[X Link](https://x.com/deredleritt3r/status/2022348084483621125)  2026-02-13T16:32Z 10.2K followers, [---] engagements


"It's already fast enough. If I give a research task to a junior lawyer I'll get the answer back in a few hours (on average) maybe even the next day. GPT-5.2 Pro gives me the answer in 30-40 minutes. In either case I have to double-check the work (probably not a material difference in time on average). But even faster is always better https://twitter.com/i/web/status/2022351855758483855 https://twitter.com/i/web/status/2022351855758483855"  
[X Link](https://x.com/deredleritt3r/status/2022351855758483855)  2026-02-13T16:47Z 10.2K followers, [----] engagements


"@tombielecki @emollick @GaryMarcus @FT Probably just easier to offload even minor research tasks to the model instead of looking things up myself"  
[X Link](https://x.com/deredleritt3r/status/2022363334176297304)  2026-02-13T17:32Z 10.2K followers, [---] engagements


"Would it be possible for you to explain to the non-technical crowd just how important this result is Would you consider this a truly novel discovery made by the harnessed version of GPT-5.2 Pro A significant truly novel discovery Or is this more similar to solutions to Erdos problems (in many cases possible to be solved by expert humans but no one bothered doing the work for whatever reason) https://twitter.com/i/web/status/2022445684578726265 https://twitter.com/i/web/status/2022445684578726265"  
[X Link](https://x.com/deredleritt3r/status/2022445684578726265)  2026-02-13T23:00Z 10.2K followers, [----] engagements


"@daniel_mac8 In fact an internal model they're "currently training""  
[X Link](https://x.com/deredleritt3r/status/2022694761283670345)  2026-02-14T15:29Z 10.2K followers, [---] engagements


"@TheRealAdamG Also Codex App for Enterprise users (in Windows) wen"  
[X Link](https://x.com/deredleritt3r/status/2018430573140693203)  2026-02-02T21:05Z [----] followers, [---] engagements


"@ShanuMathew93 The merger combining the two companies seems to have already occurred based on filings with the Nevada Secretary of State. https://x.com/AndrewCurran_/status/2018432217626935655s=20 It's happening. https://x.com/AndrewCurran_/status/2018432217626935655s=20 It's happening"  
[X Link](https://x.com/deredleritt3r/status/2018437198450770096)  2026-02-02T21:31Z [----] followers, [---] engagements


"@dylan522p Not sure whether that's true. Adding xAI shareholders to SpaceX shareholders could have diluted Google"  
[X Link](https://x.com/deredleritt3r/status/2018449459466641548)  2026-02-02T22:20Z [----] followers, [----] engagements


"@mo_ksin @AndrewCurran_ As far as I can tell SpaceX formed a subsidiary merger sub which merged with the xAI holding company. So it's an acquisition by SpaceX of xAI that was structured as including this merger"  
[X Link](https://x.com/deredleritt3r/status/2018460186697158709)  2026-02-02T23:03Z [----] followers, [---] engagements


"@molzmann1 @btibor91 100% yes"  
[X Link](https://x.com/deredleritt3r/status/2018492087289651441)  2026-02-03T01:09Z [----] followers, [--] engagements


"@dylan522p Looks like SpaceX shareholders are in fact getting diluted. Each xAI share is converting into .1433 SpaceX shares. https://x.com/MeghanBobrowsky/status/2018485910401077702s=20 New: Shares in xAI will be converted to [------] shares of SpaceX stock and current and former X/xAI employees will have the option to cash out shares in the coming weeks according to a copy of an email viewed by The Journal https://t.co/3oSmi1pg5U https://x.com/MeghanBobrowsky/status/2018485910401077702s=20 New: Shares in xAI will be converted to [------] shares of SpaceX stock and current and former X/xAI"  
[X Link](https://x.com/deredleritt3r/status/2018497277069869510)  2026-02-03T01:30Z [----] followers, [----] engagements


"@jwblackwell I already use both in the terminal. But 99.9% of lawyers won't touch it and that's why a good UI is exciting to me"  
[X Link](https://x.com/deredleritt3r/status/2018681501953491051)  2026-02-03T13:42Z [----] followers, [---] engagements


"@firstadopter @Chris_Ciaccia Not yet. But with computer use"  
[X Link](https://x.com/deredleritt3r/status/2018806128227094789)  2026-02-03T21:57Z [----] followers, [---] engagements


"@daniel_mac8 Ads are for free users and Go users only don't influence ChatGPT's responses and OpenAI keeps conversations private from advertisers. There's nothing wrong with any of this. https://x.com/cryps1s/status/2019087857642594810s=20 Answer independence: Ads do not influence the answers ChatGPT gives you. Answers are optimized based on what's most helpful to you. Ads are always separate and clearly labeled. Conversation privacy: We keep your conversations with ChatGPT private from advertisers and we never https://x.com/cryps1s/status/2019087857642594810s=20 Answer independence: Ads do"  
[X Link](https://x.com/deredleritt3r/status/2019135998664536099)  2026-02-04T19:48Z [----] followers, [---] engagements


"It's fine. All's fair in love and war. If Anthropic wants to release an ad targeting OpenAI I think that's a bit silly personally (and even pretty bad look given that the ad seems to misrepresent how OpenAI is actually implementing ads) but competitors in other industries do this kind of nonsense all the time. I agree with you that Anthropic isn't targeting the consumer market anymore and I don't think the purpose of the ad is to grab a share of that market. https://twitter.com/i/web/status/2019137845114634672 https://twitter.com/i/web/status/2019137845114634672"  
[X Link](https://x.com/deredleritt3r/status/2019137845114634672)  2026-02-04T19:55Z [----] followers, [--] engagements


"I think Free tier includes Sonnet [---] (with a daily limit) and no Opus [---]. I would rather use OpenAI's free tier with ads than Anthropic's free tier without ads - based purely on the quality of the models I can get from each. My guess is that Anthropic won't be doing ads because it's mostly catering to enterprise customers now. It's not possible to put ads in an enterprise product. https://twitter.com/i/web/status/2019145046419730762 https://twitter.com/i/web/status/2019145046419730762"  
[X Link](https://x.com/deredleritt3r/status/2019145046419730762)  2026-02-04T20:24Z [----] followers, [---] engagements


"I wouldn't trust anything that any lab says about ads including not only OpenAI but also Google and Anthropic. A promise that isn't legally binding isn't worth the paper it's written on. However I find it difficult to believe that OpenAI will shift its stance on ads materially in the future - if only because of market forces. Letting ads influence responses from ChatGPT would be a disaster (I would never use ChatGPT again). Incorporating ads into Plus and Pro tiers or enterprise plans would drive paying users away in droves. The one exception is OpenAI's current commitment not to sell data to"  
[X Link](https://x.com/deredleritt3r/status/2019149622116053259)  2026-02-04T20:42Z [----] followers, [----] engagements


"I don't think there's a way to do it. E.g. Anthropic could edit its governance documents to say "no ads ever" but what would stop Anthropic from reversing the edit at a later date I suppose Anthropic could make a voluntary commitment to some regulatory agency (maybe a state AG) but this seems overkill and unnecessary. https://twitter.com/i/web/status/2019151541396259222 https://twitter.com/i/web/status/2019151541396259222"  
[X Link](https://x.com/deredleritt3r/status/2019151541396259222)  2026-02-04T20:50Z [----] followers, [----] engagements


"@shaunralston @daniel_mac8 This is true but at the same time it's pretty obvious which lab this ad was really about"  
[X Link](https://x.com/deredleritt3r/status/2019164947012600188)  2026-02-04T21:43Z [----] followers, [--] engagements


"@bazirani01 As far as I can tell Google is pursuing a different path. Which by the way I am also thankful for - I think two separate bets on how to reach AGI by different labs is the best outcome for the general public"  
[X Link](https://x.com/deredleritt3r/status/2019166694850064429)  2026-02-04T21:50Z [----] followers, [---] engagements


"I don't have a good tool for contract edits right now but I'm guessing that will change soon. There are ways to do it with Claude Code/Codex but because lawyers are extremely non-technical we need a nice UI to be served to us on a silver platter. I think Anthropic recently released a legal tool (haven't tested it) and providers like Harvey may have something also(). Like everything else with AI this is all developing very fast and quite hard to keep track of. https://twitter.com/i/web/status/2019184880177872918 https://twitter.com/i/web/status/2019184880177872918"  
[X Link](https://x.com/deredleritt3r/status/2019184880177872918)  2026-02-04T23:02Z [----] followers, [---] engagements


"Sam Altman: "Very little discussed are the hardware implications of recursive self-improvement: robots that can build other robots data centers that can build other data centers chips that can design their own next generation. Maybe the problem of chip design will turn out to be a very good problem for previous generations of chips." My excellent Conversation with Sam Altman: https://t.co/YkUAJnglJm @sama My excellent Conversation with Sam Altman: https://t.co/YkUAJnglJm @sama"  
[X Link](https://x.com/deredleritt3r/status/1986115691150233789)  2025-11-05T16:57Z 10.2K followers, 49.9K engagements


"Dear "AI bubble will pop" doomers I hate to break it to you but: - if the bubble pops tomorrow and OpenAI/Anthropic go bankrupt their assets (including the frontier models and datacenter assets/rights) will just be acquired on the cheap by the big tech companies. Microsoft will continue OpenAI's mission. Amazon will continue Anthropic's mission. Google will just be Google. - when a company goes bankrupt its key personnel don't just magically evaporate. The best researchers the model IP and the compute will quickly find each other again albeit in slightly new teams and under a new corporate"  
[X Link](https://x.com/deredleritt3r/status/2013106314654720316)  2026-01-19T04:28Z 10.2K followers, 237.8K 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 [---] Deep Think Claude Opus [--] (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/anyuser/status/1974862963442868228)  2025-10-05T15:43Z 10.2K followers, 221.2K engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing