Skip to content

Midnight Update · Nov. 21

A text from Dad: what’s happening in ai rn? The reply: Stanford Mathematician Claims a Human-Only Riemann Proof

The boy says: A HUMAN PROVED THE SAME THING ALONE.

The boy from Ai Lately's Sunday Funnies, in glasses, a white T-shirt and black shorts: both hands up, brows high, a small round mouth.

Saturday, Nov. 21, 2026 · Midnight Update

Stanford Mathematician Claims a Human-Only Riemann Proof

A Stanford mathematician says he proved the Riemann Hypothesis alone, days after an Anthropic researcher claimed it with Claude.

Number 1, Viral on X

Thomas Bloom @thomasfbloom

Nov. 20 · 1:37 p.m. ET

Stanford Mathematician Claims Human-Only Proof of Riemann Hypothesis

@thomasfbloom wroteAn independent, 100% human, claimed proof of RH (only 940 pages long).
5,870
Likes
On X
669
Reposts
On X
106
Replies
On X

The takeaway

A Stanford mathematician claims a human-only proof of the Riemann Hypothesis, days after Anthropic's AI-assisted claim.

Number 2, Viral on X

Anshuman Singh @anshuman_ssh

Nov. 20 · 1:37 p.m. ET

Arcfront Launches With $1 Billion Run Rate

@anshuman_ssh wroteIn under a year, four of the five frontier labs became customers.
1,359
Likes
On X
194
Reposts
On X
203
Replies
On X

The takeaway

A tiny startup says it reached a $1 billion run rate on under $2 million raised.

Number 3, Lab post · RadixArk

RadixArk @radixark

Nov. 18 · 1:04 p.m. ET

RadixArk Ships DaisyRose, a 27-Billion Parameter Model

@radixark wroteHow much can you improve a small model in 18 days on 144 GPUs?
817
Likes
On X
60
Replies and quotes
On X
277
Hugging Face likes
Beyond X

The takeaway

A small lab trained an open model for 18 days and doubled a coding benchmark score.

You're caught up

That's AI right now.

The posts that drew the most attention in the 6 hours before 12:07 a.m. ET.

Best ten in the markets

Gain since the previous close · Nov 20, 11:29 PM ET · Yahoo Finance

  1. 1iShares 20+ Year Treasury TLT · Bond+1.37%
  2. 2iShares 7-10 Year Treasury IEF · Bond+0.91%
  3. 3iShares Core U.S. Aggregate AGG · Bond+0.62%
  4. 4iShares Investment Grade Corporate LQD · Bond+0.59%
  5. 5Vanguard Total Bond Market BND · Bond+0.59%
  6. 6iShares National Muni Bond MUB · Bond+0.40%
  7. 7iShares 1-3 Year Treasury SHY · Bond+0.28%
  8. 8IBM IBM · Stock+0.25%
  9. 9iShares High Yield Corporate HYG · Bond−0.26%
  10. 10SPDR High Yield Bond JNK · Bond−0.34%

Charts for each, in the Money columnA record of price, not investment advice.

A text from Dad: what’s happening in ai rn? The reply: Stanford Mathematician Claims a Human-Only Riemann Proof

Six times a day: what blew up in AI on X and Reddit, why it caught fire, and what it says about the people building AI.

The boy says: A HUMAN PROVED THE SAME THING ALONE.

The boy from Ai Lately's Sunday Funnies, in glasses, a white T-shirt and black shorts: both hands up, brows high, a small round mouth.

Midnight Update, 12:07 a.m. ETCovers 6:07 p.m. Fri. to 12:07 a.m. Sat. ET

An earlier update. Read the latest

Stanford Mathematician Claims a Human-Only Riemann Proof

A Stanford mathematician says he proved the Riemann Hypothesis alone, days after an Anthropic researcher claimed it with Claude.

Published 12:07 a.m. ET by Ai Lately

Next update in--:--:--Four updates a day, six hours apart

AI RN shows each X post as plain text. X's script loads only when you ask.

Number 1

Viral on X

Stanford Mathematician Claims Human-Only Proof of Riemann Hypothesis

Thomas Bloom @thomasfbloom on XPosted

Thomas Bloom@thomasfbloom

An independent, 100% human, claimed proof of RH (only 940 pages long). (Seems like this was finished months ago, but has been privately being reviewed.)https://t.co/xz4sFkNesk

Open on X (opens a new tab)

By the numbers

Likes
5,870
On X
As of
Likes: Source: X public data
Reposts
669
On X
As of
Reposts: Source: X public data
Replies
106
On X
As of
Replies: Source: X public data
Hacker News points
1,780
Beyond X
As of
Hacker News points: Source: Hacker News
3 more numbers behind this take

The take

A Stanford mathematician says he proved the Riemann Hypothesis. He worked alone. He skipped any AI help.

His name is Alessandro Masullo. He spent twelve years on the proof. He wrote it down in 940 pages.

The Riemann Hypothesis is an old math puzzle. It is about how prime numbers are spaced out. Mathematicians have chased it for 167 years.

Masullo's timing stunned readers. Anthropic researcher Levent Alpöge had said days earlier that he solved the same puzzle. Alpöge worked with Claude, Anthropic's AI model. Claude can write and check math proofs too.

Brian Conrad is a math professor at Stanford. He read drafts of Masullo's proof for months. Mathematician Elliot Glazer said Conrad found the draft looking good so far.

Masullo wrote his own note about the moment. "We are still here," he wrote. "We are very much alive. We are not going anywhere."

Science writer Kevin Hartnett shared the news on X soon after. His post earned 9,435 likes within hours.

Why did this catch fire? Two proofs of the same old puzzle landed days apart. One came from a person working with AI. The other came from a person working alone for over a decade. That split put one question in front of every reader. Can a person still do this kind of work on their own?

The post sharing Masullo's claim drew 5,870 likes on X. It earned 669 reposts and 106 replies. A linked write-up on Hacker News earned 1,780 points. Readers left 800 comments there.

What does this say about the people who build AI? Labs now race to claim hard math wins first, and fast. Masullo's twelve quiet years show a slower pace still exists, and still works. Mathematicians will check both proofs for months to come. Only then will anyone know which one truly holds.

What it says about the people in AI

A Stanford mathematician claims a human-only proof of the Riemann Hypothesis, days after Anthropic's AI-assisted claim.

Sources

  1. TechCrunch AI claimed a solution to the 167-year-old Riemann hypothesis — and a human did, too
  2. Scientific American Dueling claims of Riemann hypothesis proofs emerge

Number 2

Viral on X

Arcfront Launches With $1 Billion Run Rate

Anshuman Singh @anshuman_ssh on XPosted

Anshuman Singh@anshuman_ssh

Today we're launching @arcfront with @Naman_Bhalla and deep support from @asxna.

In under a year, four of the five frontier labs became customers.

In 2025 an AI solved all 12 problems at the ICPC World Finals. The best human team solved 11. I competed in two of those finals,… pic.twitter.com/KbqGFW5Z4j

Open on X (opens a new tab)

By the numbers

Likes
1,359
On X
As of
Likes: Source: X public data
Reposts
194
On X
As of
Reposts: Source: X public data
Replies
203
On X
As of
Replies: Source: X public data
Enterprise partners
300+
Beyond X
As of
Enterprise partners: Source: cellcog.ai
1 more number behind this take

The take

A startup called Arcfront came out of hiding today. Its founder is Anshuman Singh.

Arcfront says it reached a $1 billion run rate. A run rate means yearly sales, guessed from recent months. Singh says the company raised under $2 million to get there.

Arcfront builds training data for AI labs. That data shows how human experts make hard choices. Labs use it to teach their models good judgment.

Four of the five top AI labs now pay for this data. Arcfront also counts over 300 business partners. The company built more than 1,200 tasks pulled from real jobs. Those jobs span health care, law, and banking.

Singh pointed to a math contest as proof of AI's new reach. In 2025, an AI solved all twelve problems at the ICPC World Finals. The best human team solved eleven. Singh said he competed in two of those contests himself.

Why did this catch fire? A tiny team claimed huge sales on barely any funding. That ratio stunned readers used to AI startups burning cash fast. Arcfront's pitch also ties plain data work to a famous AI win at a math contest.

Singh's launch post earned 1,359 likes on X. It drew 194 reposts and 203 replies. Outside coverage named Arcfront's run rate too, alongside its count of over 300 enterprise partners.

What does this say about the people who build AI? Frontier labs now pay smaller firms to teach their models judgment, on top of plain facts. A founder who once competed in coding contests now sells proof that AI beat his old rivals. Expert judgment, it turns out, carries a real price for the labs racing ahead. Money keeps moving toward whoever can package human skill into data fastest.

What it says about the people in AI

A tiny startup says it reached a $1 billion run rate on under $2 million raised.

Sources

  1. Cellcog Arcfront Launches: Expert Data for 4 of 5 Frontier Labs

Number 3

Lab postRadixArk

RadixArk Ships DaisyRose, a 27-Billion Parameter Model

RadixArk @radixark on XPosted

RadixArk@radixark

How much can you improve a small model in 18 days on 144 GPUs?

Meet DaisyRose-27B, RadixArk's first open-weight model! It is trained with RL on Qwen3.8-27B, with 1,200 curated long-horizon agent tasks for 18 days on 144 H200s. Powered by Miles and our own training backend, we… pic.twitter.com/y7b8A9K5av

Open on X (opens a new tab)

By the numbers

Likes
817
On X
As of
Likes: Source: X public data
Replies and quotes
60
On X
As of
Replies and quotes: Source: X public data
Hugging Face likes
277
Beyond X
As of
Hugging Face likes: Source: Hugging Face
Hugging Face downloads
918
Beyond X
As of
Hugging Face downloads: Source: Hugging Face

The take

RadixArk just shipped its first open model. It is called DaisyRose-27B.

RadixArk is a small AI lab. DaisyRose-27B has 27 billion parameters. A parameter is a number stored inside a model. More parameters can mean more skill, but they also cost more to run.

The team built DaisyRose in 18 days. They used 144 computer chips called H200s. Those chips train AI models fast.

DaisyRose learned from 1,200 real work tasks. These were long tasks, like typing commands into a terminal. The model learned through reinforcement learning. That means it earned rewards for right answers and learned from its own mistakes.

The results were strong. DaisyRose more than doubled its base model's score on Terminal-Bench 4.1. That score jumped from 5.6 to 12.0. Terminal-Bench 4.1 tests how well a model can run a command line on its own.

The model carries an Apache 2.0 license. That license lets anyone use, change, and share the model free of charge.

Why did this catch fire? Small labs rarely beat the giants on hard coding tests. RadixArk did it with just 144 chips. Big labs often use thousands of chips for one model. That gap made RadixArk's result land as a genuine upset.

The launch post earned 817 likes on X. It drew 60 replies and quotes. On Hugging Face, where anyone can download the model, DaisyRose earned 277 likes and 918 downloads already.

What does this say about the people who build AI? Small teams can still surprise the field. A clear task list and eighteen days beat raw size here. Coders now pull DaisyRose onto their own machines to test it. The real test comes next, once regular coders try it at their own daily work.

What it says about the people in AI

A small lab trained an open model for 18 days and doubled a coding benchmark score.

Sources

  1. DEV Community RadixArk opens DaisyRose-27B, an agent-tuned Qwen model
  2. RadixArk DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency

That is AI right now.

The AI posts and threads that drew the most attention in the 6 hours before 12:07 a.m. ET. Newer updates wait in the archive.

Next update in--:--:--Four updates a day, six hours apart

Best ten in the markets

Gain since the previous close · Nov 20, 11:29 PM ET · Yahoo Finance

  1. 1iShares 20+ Year Treasury TLT · Bond$81.45+1.37%
  2. 2iShares 7-10 Year Treasury IEF · Bond$91.65+0.91%
  3. 3iShares Core U.S. Aggregate AGG · Bond$96.15+0.62%
  4. 4iShares Investment Grade Corporate LQD · Bond$103.69+0.59%
  5. 5Vanguard Total Bond Market BND · Bond$71.33+0.59%
  6. 6iShares National Muni Bond MUB · Bond$103.51+0.40%
  7. 7iShares 1-3 Year Treasury SHY · Bond$81.61+0.28%
  8. 8IBM IBM · Stock$224.44+0.25%
  9. 9iShares High Yield Corporate HYG · Bond$76.33−0.26%
  10. 10SPDR High Yield Bond JNK · Bond$91.54−0.34%

See the charts in MoneyA record of price, not investment advice.

More updates

All updates
  1. Saturday, Nov. 21, 2026

  2. Friday, Nov. 20, 2026

  3. Thursday, Nov. 19, 2026

Keyboard

j k
Move through a list
Enter
Open the selected piece
/
Search, and browse the columns
g then r
Right Now
g then a
Articles
g then s
The Signal
g then o
Opinion
g then b
Analysis
g then p
People
?
This sheet
Esc
Close

AI Right Now is free with an account.The Citation Audit is free with an account.Skills are free with an account.Tutorials are free with an account.You have read your 3 free articles.

Every article and comic, AI Right Now, the Citation Audit, Skills and Tutorials.

Join Ai Lately for free

Join with Google or your email. Already a member? Sign in

Press anywhere else for the home page.