Founder of a stealth agentic operating system company TBA soon™️- I might not always be right but I'm never wrong
Joined September 2013
- Tweets36.8K
- Following703
- Followers2.7K
- Likes40.5K
The Merstradamous 🇦🇱🇺🇲 retweeted
Anthropic Subscriptions Offer 5x+ More Value Than OpenAI
Limit testing every AI subscription plan from Anthropic, OpenAI, Meta, SpaceXAI, MiniMax, Moonshot, Zdotai, Cursor, and Cognition
newsletter.semianalysis.com/…
The Merstradamous 🇦🇱🇺🇲 retweeted
A 37-year-old man with a rare blood disease was being sent to hospice.
An AI model searched thousands of existing drugs and ranked a combination no one had tried for his condition.
He responded within a week. He is in remission.
The drugs were already on the shelf. The AI model found the match.
The Merstradamous 🇦🇱🇺🇲 retweeted
Since the Codex reset yesterday, I already depleted the weekly limits of one of my Codex accounts.
This account exlusively uses GPT-5.6 Sol. I'm tracking all token consumption through my OpenClaw setup, so it's very easy to compare to previous periods.
Limits are now exhausting 4.8-5.9x faster than they did just 2 months ago while consumption is now around ~18% slower.
Same reasoning efforts, same mixed type of work.
The token rug pull is coming.
Get ready to pay up.
We have a problem!
There's a huge token rug pull coming.
Right now, Claude and Codex max subscriptions are SUPER generous, and it won't last.
We've seen it before:
- First, a huge AI credit giveaway.
- Then the models get dumber.
- Credits start running out faster.
- Prices jump / the offer ends.
If you rebuilt your company to be entirely dependent on OpenAI's / Anthropic's generosity, you're in for a bad surprise soon 👇
The Merstradamous 🇦🇱🇺🇲 retweeted
Agreed. Intelligence per joule has accelerated sharply—around 18x in 16 months from architectures and hardware, with algorithmic gains compounding roughly 3x yearly. That multiplies any energy-scale advantage and compresses timelines beyond static compute numbers. US high-end lead holds today, but the rate favors whoever scales energy and efficiency fastest.
The Merstradamous 🇦🇱🇺🇲 retweeted
1/ Today, we’re introducing Geodesic Intelligence, and launching NovaDDE and NovaAtom-Lite-Preview.
Geodesic is building an AI-native platform for protein therapeutics, building AGI for drug discovery to find the shortest path from biology to medicines.
geodesiclab.com
The Merstradamous 🇦🇱🇺🇲 retweeted
This is one of the more interesting agent architectures I’ve seen lately.
Instead of bolting persistence, identity, coordination, and recovery onto an agent loop, Gents flips the stack: the database becomes the control plane and agents become durable, distributed processes over it.
The future of agents probably looks a lot less like "LLM + tool loop" and a lot more like systems that can persist, coordinate, recover, and move across machines without losing themselves.
github.com/gents-ai/gents
The Merstradamous 🇦🇱🇺🇲 retweeted
Knowledge workers are starting to sound exactly like artists did when generative Al arrived:
"Real professionals won't use this."
"It can't replace actual skill."
"It's just a toy."
When you've spent your entire life mastering a scarce skill, it's hard to accept that the scarcity itself is disappearing.
Expertise isn't becoming worthless. But the moat around expertise is collapsing.
Al is giving ordinary people access to capabilities that once took years or decades to acquire.
That's the part people are still underestimating
Many people have no idea what's happening.
They can't conceive that developers out there are deploying AI-generated code, because in their bubble, everyone is still writing Java code and running Scrum retrospectives.
One sunny afternoon, their CTO will meet a modern developer at church; they will talk about the possibilities and how AI can change their business. A month later, these dinosaurs will face a tough decision: either get up with the times, or find another job.
The Merstradamous 🇦🇱🇺🇲 retweeted
meet @mostik_ai!
what happens when you put 12 PhDs in one room for four months? first place on the ARC-AGI leaderboard, which I can't say much about while the competition is still running. and this, which I can.
everyone's arguing about whether open models will catch up to frontier models. we think it's the wrong question. here's the one we pose: why does a frontier model have to generate your answer at all, when the only thing you need from it is the reasoning?
we do this by enabling models to communicate in latent space. through our protocol, hidden states pass straight from a frontier model into a small one running on your infrastructure -- no text between them, and neither model is fine-tuned. two models from different families, sharing reasoning, both left untouched.
how do we know it works? we tested it on a setup where a 753B model reads the problem, and a 4B edge-class model writes the answer. with this approach, we get results 80% as accurate as the frontier model, but at 20x faster performance.
we're committed to preventing frontier model lock-in and are already partnering with inference providers to accelerate open-weight adoption. we've done this between 15 of us, in four months, 12 PhDs and a Fields medalist, backed by @generalcatalyst
WIRED has the first external account of the company and the work: wired.com/story/russian-star…
full writeup, the setup, and all the numbers: mostik.ai/read-more
The Merstradamous 🇦🇱🇺🇲 retweeted
这次 ChatGPT quota nerf，基本已经被社区实锤了，下降了将近 76% 🤬
SemiAnalysis 6 月曾实际测试过各档 ChatGPT 订阅，当时估算出来的 API 等值大约是：
$20 Plus → ~$700/月
$100 Pro 5x → ~$3,500/月
$200 Pro 20x → ~$14,000/月
但最近 Reddit 上几位用户测出来的数据是：
Plus：额度下降约 20%～50%
Pro：部分账户下降 70%+ ！！！
其中一个 Plus 用户的 NerfTrack 数据尤其有意思。
上周 reset 之后，他观察到的并不是“一刀砍到底”：
OpenAI 第一次先砍约 40%，发现没人注意后，又继续砍了 10%，最终相对原额度只剩一半。
另一个 Pro 5x 用户的数据更夸张：
$674/周 → $157/周
直接缩水约 76.7%。
所以 Pro 用户多付 5倍的价格结果却拿到跟 Plus 差不多一样的额度 ？？？
OpenAI 你是真的把用户当猴耍啊 🤬
Tibo 搞不好又要用是 bug 搪塞过去了 🙄
如果你最近感觉 Codex “怎么突然这么容易花完 quota”，你可能真的不是错觉。
Plus / Pro 用户来报个数：
你现在一周实际能跑多少？
我尤其想看看 Pro 用户的数据 👀
SemiAnalysis：
teapawt.org/i/status/2064815044085…
Reddit：
reddit.com/r/codex/comments/…
reddit.com/r/codex/comments/…
Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It's widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found that the subscriptions are actually far more generous. (2/4)
The Merstradamous 🇦🇱🇺🇲 retweeted
looks like Codex limits may have been quietly nerfed.
back in June, SemiAnalysis tested the ChatGPT plans and estimated their max API-equivalent usage at roughly:
$20 Plus → ~$700/month
$100 Pro 5x → ~$3,500/month
$200 Pro 20x → ~$14,000/month
now a Plus user tracking with NerfTrack says their weekly limit dropped from around $160 → $80, basically a 50% cut.
one Pro 5x user also shared numbers showing $674 → ~$157/week, which is around a 77% drop.
a lot of people have been noticing their Codex quota disappearing much faster lately too.
hopefully this is just a bug, right @thsottiaux ? right??😭
The Merstradamous 🇦🇱🇺🇲 retweeted
🚨 Reddit users say OpenAI may have cut Codex limits in half
NerfTrack estimates the weekly ChatGPT Plus limit dropped from roughly $160 in API value to about $80 within a few days. Users on the $100 Pro 5x and $200 Pro 20x plans are also reporting their weekly quota now disappears in 1 to 3 days.
reddit.com/r/codex/comments/…
The Merstradamous 🇦🇱🇺🇲 retweeted
Mark Zuckerberg gets it right:
“The defining question of our age isn’t whether superintelligence will exist, but who will have access to it. Will it be centralized and restricted to a few institutions, or will it be a tool that empowers everyone?”
Concentration of power is the biggest risk of AI. When a small number of labs (working hand-in-glove with the administrative state) decide who has access to which model capabilities, they inevitably shape what can be said, known, and built. That’s not “safety.” It’s control.
As Mark points out, the history of open source shows that broad access and transparency are usually the best path to actual security and resilience. Decentralization creates checks and balances on power. By contrast, centralized alternatives, like bureaucratic approval regimes and mandatory gatekeeping, typically produce regulatory capture and reinforce cartels.
Personal superintelligence in everyone’s hands, with competing models and real data sovereignty, is a far better check on a dystopian future than self-appointed guardians who claim to be “aligned” with all of humanity.
The Merstradamous 🇦🇱🇺🇲 retweeted
Anthropic maintains that it is entitled to train for free on all the world’s output, even if the author objects. But if a competitor trains on Anthropic’s output after paying for it, that is IP theft. The hypocrisy is breathtaking.
btw anthropic's internal document on this literally said "we don't want it to be known that we are working on this.”
it was called project panama.
here's exactly what happened:
1: anthropic concluded that books were the cheapest way to build a world-class model because they gave claude curated facts, structured arguments, compelling stories, and writing “an editor would approve of.”
2: once anthropic decided it needed books at enormous scale, its first solution was piracy.
it downloaded 7m+ books from online libraries including libgen. the judge later wrote that although anthropic had legal ways to buy them, it chose piracy to avoid what dario amodei called the “legal/practice/business slog.”
3: that piracy created a massive legal risk.
so in february 2024, anthropic hired tom turvey, the former head of partnerships for google books, to find a legally safer way of obtaining “all the books in the world.”
4: turvey first contacted major publishers about licensing their catalogs.
those attempts didn’t produce agreements, so anthropic chose a route that required no publisher permission: buying millions of physical books through distributors and used-book retailers.
5: within about a year, anthropic spent tens of millions acquiring and scanning millions of books, including many rare and 1/1 titles. one vendor proposal targeted 500,000 to 2 million books in six months.
6: to scan that many books within months, the vendors physically dismantled them.
a hydraulic cutter removed each spine. the pages were trimmed to size, fed as loose sheets through high-speed industrial scanners, and converted into searchable PDFs. the paper remains were then sent for recycling.
7: these PDFs were fed into claude as training data.
the complete collection became a private, searchable anthropic library that the company planned to “store forever.” the scans aren’t available to the public and were never open-sourced.
The Merstradamous 🇦🇱🇺🇲 retweeted
Kimi made it to Hugging Face and is now open weights, with a paper too!
Releasing the model weights and technical report of Kimi K3.
Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.
New model architecture: 2.5x the intelligence per unit of compute, not just more params.
Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.
Model weights: huggingface.co/moonshotai/Ki…
Tech report: github.com/MoonshotAI/Kimi-K…
Tech blog: kimi.com/blog/kimi-k3
The Merstradamous 🇦🇱🇺🇲 retweeted
David Sacks: Anthropic Is Trying to Crush Open Source AI and the American Developers Who Use It
@DavidSacks:
“I know people don't have a lot of sympathy for Chinese companies, that's fine. I'm not defending Chinese companies.
I'm defending American developers who need to be able to use everything in the public domain.
And let me give you an example. Cursor rolled out its new product, Composer 2. They were able to post-train that model using Kimi K2.5 on their own proprietary coding data.
They started with a Chinese open source model, and then they used their own data, and they came up with a new derivative product.
This is the way that open source works. You take things that are in the public domain, you fork them, you make them your own.
And by the way, once it's in the public domain, it's not a Chinese model anymore. No data is going back to China, nothing's going back to China.
An American company has taken open source contributions in the public domain, made it their own, and then developed their own model.
And if you say that American companies can't do that, or that somehow it's tainted with IP theft, you are basically going to put a dagger through the heart of the entire American open source ecosystem.
And that is exactly what Anthropic wants, because they do not want to have the competition.”
The Merstradamous 🇦🇱🇺🇲 retweeted
At the $200 tier, Claude Code is far more generous than Codex right now.
Sol can burn through your usage in a day or two.
Opus 5 is so token-efficient that I can run 4–5 Claude Code sessions daily without hitting the limit.
I’d pick Claude Code any day.
The Merstradamous 🇦🇱🇺🇲 retweeted
Up until now, I didn’t think a Claude subscription was worth it over GPT because of the usage limits.
Now, Claude offers a more powerful model and way higher usage limits.
It doesn't matter how many times they throw out extra resets—that’s just temporary rations.
The Merstradamous 🇦🇱🇺🇲 retweeted
Glad everyone else is noticing that Codex usage limits have been destroyed.
>It’s draining as fast as 5hr limit. But weekly
>Cheaper models, are seeming to drain our quota just as fast.
>20x plans feel like 5x
“5.6 is extremely efficient, and costs just as much as 5.5” yeahh right.
Sorry but Codex Usage limits are terrible now?
Yesterday, i only used 5.6 sol high, i let it running in a /goal, configured cheaper terra/ lunar subagents.
I come back 40% of my total weekly usage is gone? ( 20x)
BTW i was using multiple 5.6 Sol xhigh + Max, all day everyday, last week.
Id rather get no resets, steady and fair usage, as opposed to getting resets, but every time they do, they reduce our quota.
They also removed the invite friends for banked resets.
Anyone else noticing this? Just gets worse by the day.
The Merstradamous 🇦🇱🇺🇲 retweeted
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-W…
The Merstradamous 🇦🇱🇺🇲 retweeted
If your Codex usage is draining down quickly since last one week then you are not alone.
Codex is purposely doing this.
100s of customers have reported the issue but no word from @OpenAIDevs @OpenAI's team so far.
@thsottiaux please look into this and update us.
