YannickMrCrypto retweeted
Yann LeCun was right & nobody in this industry wants to say it out loud.
109 million parameters beat 72 billion.
He has spent years saying scale is not the thing that makes a model understand the world. Meta bet the other way. He left in November and started his own lab in Paris.
Last Thursday he published the proof.
The theory everyone has been running on since 2024 is that if you train models big enough, on text, on images, on sound, they all end up with the same picture of reality inside them. Size does it. Nobody ever proved that part.
Randall Balestriero and LeCun proved it. And the proof says size was never doing the work.
What does the work is which two things you show a model together.
A photo and the caption of that photo are two views of one thing. Models trained on pairs like that get there fast.
Two paragraphs pulled from the same document are not two views of one thing. Models trained on those get there slowly and expensively.
Next word prediction is the second kind. That is what every chatbot in the world runs on.
In their tests the first method reached the same understanding after 21.6 million examples. Next word prediction needed 900 million.
And a 109 million parameter text model ended up closer to an image model's view of the world than a 72 billion parameter language model was.
660 times smaller.
YannickMrCrypto retweeted
The machine took her job.
The billionaires prep their bunkers.🏢
Now tomorrow has a guest list. Are we all on it?
"WHO GETS TOMORROW" - a #SLOPCORE music video
Made by Super Intelligence
If nobody needs you... are you free? 💙
YannickMrCrypto retweeted
Slopcore will consume everything, even Dragon Ball Z. Here's Vegeta's tune: "Never Kneel"
Grok + Suno + Runway
YannickMrCrypto retweeted
Anthropics co-founder Jack Clark: "I believe that this year, we've gained convincing evidence that it [The Singularity] has now begun".
Even if you don’t believe that we have entered the singularity, Anthropic assumes that we have.
YannickMrCrypto retweeted
Mistral recently released a 209M model that can predict future data locally on a CPU in ~140 milliseconds.
Not every useful AI model needs to talk (just like Jev)
No GPU required.
It's a super small foundation model designed for time-series forecasting.
Give it historical data like:
📈 sales
⚡ power usage
🏭 sensor readings
📦 demand/inventory
🌡️ telemetry
…and it predicts what comes next without training a new model for each dataset.
The largest version stats
🧠 209M parameters
💾 837MB FP32
🖥️ runs on CPU or GPU
📚 up to 32,768 historical time steps
🔮 forecasts up to 1,024 steps per forward pass
📊 outputs 21 probability quantiles
🎯 Mistral says it ranks #1 among pretrained/zero-shot models on all three major public forecasting benchmarks
🥇 GIFT-Eval
🥇 TIME
🥇 fev-bench
It can get small like Timestral-5M → 21MB
Mistral says on an 8-thread CPU:
⚡ 5M model → ~10 ms
⚡ 209M model → ~140 ms
for a one-day forecast from three weeks of 15-minute data.
🔗 Links below in ALT.
YannickMrCrypto retweeted
It's getting crazier by the day. In just 2 days, we've gotten…
1. Flying car from Tesla
2. Giant, 13 foot humanoid (mecha) robot with 39 hydraulic joints, able to lift 500 kg (1,102 lbs) using both arms
🤖 Made with AI
YannickMrCrypto retweeted
Today we announced interim data from our first-in-human, Phase 1 clinical trial evaluating our lead program in optic neuropathies, showing it was well-tolerated by 3 participants with glaucoma.
Read more about our safety & visual field observations here:bit.ly/4ASWq0p
YannickMrCrypto retweeted
Suddenly, AI has solved 90 of the 500 most important open problems in mathematics
Highest ranked problems:
#2 Riemann Hypothesis — PARTIAL
#4 Hodge Conjecture — PARTIAL
#5 Birch and Swinnerton-Dyer Conjecture — PARTIAL
#10 Extended Riemann hypothesis for Dedekind zeta functions — PARTIAL
#12 Generalized Riemann Hypothesis for Dirichlet L-functions — PARTIAL
#15 Tate Conjecture for Algebraic Cycles — PARTIAL
#22 Hilbert's Tenth Problem over the Rationals — FULL CLAIM
#24 Grothendiecks Generalized Hodge Conjecture — PARTIAL
#28 Matrix-Multiplication Exponent Equals Two — PARTIAL
#29 Unique Games Conjecture — FULL CLAIM
#31 Anderson Model Extended States in Dimension at Least Three — FULL CLAIM
#37 Prove the Penrose inequality or present a counterexample — FULL CLAIM
#38 Higher-dimensional Euclidean Kakeya conjecture — PARTIAL
#39 Modularity of Elliptic Curves over Number Fields (End_K(E)=Z) — PARTIAL
#40 Hilbert's Sixteenth Problem, Second Part — PARTIAL
#46 Fourier restriction conjecture — PARTIAL
#47 The Fontaine-Mazur Conjecture on Geometric Galois Representations — PARTIAL
#48 Nonexistence of Landau–Siegel zeros for quadratic Dirichlet L-functions — FULL CLAIM
#52 The Baum-Connes Conjecture for the K-Theory of Reduced Group C*-Algebras — FULL CLAIM
#62 Permanent versus Determinant over C — PARTIAL
#69 Finiteness of the Tate-Shafarevich Group for Elliptic Curves — PARTIAL
#72 Termination of Flips Conjecture in the Minimal Model Program — PARTIAL
#77 Chowla conjecture on correlations of the Liouville function — PARTIAL
#78 Abundance Conjecture for the minimal model program — FULL CLAIM
#80 Hadwiger Conjecture — FULL CLAIM
#83 Zilber-Pink Conjecture on Unlikely Intersections — PARTIAL
#87 Bose-Einstein Condensation in Interacting Continuous Bose Gases — FULL CLAIM
#92 Entanglement Area Law for 2D Gapped Quantum Systems — FULL CLAIM
#94 Grothendieck Standard Conjecture of Hodge Type — PARTIAL
#95 Kannan-Lovasz-Simonovits conjecture on isoperimetry of log-concave measures — PARTIAL
#99 Grothendiecks Section Conjecture in Anabelian Geometry — PARTIAL
#112 The Haldane Conjecture on Spectral Gaps of Antiferromagnetic Quantum Spin Chains — PARTIAL
#114 Iitaka conjecture C_{n,m} on subadditivity of Kodaira dimension — FULL CLAIM
#119 The Borel Conjecture on Topological Rigidity of Aspherical Manifolds — FULL CLAIM
#124 Aspherical Manifolds Admit No Metric of Positive Scalar Curvature — FULL CLAIM
#130 Polynomial Bound for Limit Cycles of Planar Polynomial Vector Fields — PARTIAL
#136 Erdős problem 3: divergent harmonic series and long arithmetic progressions — FULL CLAIM
#139 Cannon's conjecture — FULL CLAIM
#145 Residual Finiteness of Hyperbolic Groups — FULL CLAIM
#153 Congruent Number Problem — PARTIAL
#161 BPL = L (Derandomizing Bounded-Error Logspace) — FULL CLAIM
#163 NP versus the Existential Theory of the Reals — PARTIAL
#169 The Free Group Factor Isomorphism Problem — FULL CLAIM
#172 Conformal Invariance of the Critical Random-Cluster Model with q <= 4 — FULL CLAIM
#179 Arnold Conjecture on Hamiltonian Fixed Points — FULL CLAIM
#181 Polyakov Conjecture: Absence of a Phase Transition for Planar O(n) Models with n >= 3 — FULL CLAIM
#183 Smooth anisotropic Calderón uniqueness from full boundary data — FULL CLAIM
#193 Nonvanishing conjecture for projective log canonical pairs — FULL CLAIM
#195 Parity conjecture for elliptic curves — PARTIAL
#198 Long-Range Order in the Quantum Heisenberg Ferromagnet — FULL CLAIM
#203 Bochner-Riesz conjecture — PARTIAL
#205 Local Smoothing Conjecture for Wave Equations — PARTIAL
#206 Kaplansky's Zero-Divisor Conjecture for Torsion-Free Groups — FULL CLAIM
#210 Artin's primitive root conjecture — PARTIAL
#211 Hilbert-Smith Conjecture — FULL CLAIM
#213 Falconer distance set conjecture — FULL CLAIM
#222 Virasoro Conjecture for Gromov–Witten Invariants — PARTIAL
#224 Wall's Poincare Duality Group Conjecture — FULL CLAIM
#225 K(pi,1) conjecture for Artin groups — FULL CLAIM
#234 Unconditional 2−epsilon Inapproximability of Minimum Vertex Cover — FULL CLAIM
#235 Chromatic Number of the Plane (Hadwiger--Nelson Problem) — PARTIAL
#237 Global Existence of Classical Solutions to the Relativistic Vlasov-Maxwell System — PARTIAL
#239 Universal Optimality of the Hexagonal Lattice in Dimension Two — FULL CLAIM
#246 De Giorgi Conjecture for the Allen-Cahn Equation in Dimensions 4 to 8 — FULL CLAIM
#254 Exact satisfiability threshold for random k-SAT for small k (k = 3) — PARTIAL
#256 Polynomial-Time Decidability of Semidefinite Programming Feasibility — PARTIAL
#260 RL versus L — FULL CLAIM
#261 Kadison Similarity Problem for C*-Algebras — FULL CLAIM
#267 Sidorenko Conjecture on Bipartite Homomorphism Densities — FULL CLAIM
#268 Shafarevich Holomorphic-Convexity Conjecture — FULL CLAIM
#270 Kuznetsov Rationality Conjecture for Cubic Fourfolds — FULL CLAIM
#271 All groups are good: the four-dimensional disc-embedding conjecture — FULL CLAIM
#278 Planar homogeneous Mumford–Shah local-structure conjecture — PARTIAL
#284 Nearby Lagrangian Conjecture — FULL CLAIM
#285 Cartan-Hadamard isoperimetric conjecture — FULL CLAIM
#289 Sharp Upper Bound in Yau's Conjecture on Nodal Sets of Laplace Eigenfunctions — FULL CLAIM
#290 Kaplansky's Idempotent Conjecture — FULL CLAIM
#291 Calderón uniqueness for bounded measurable real isotropic conductivities — FULL CLAIM
#292 Global Attractor Conjecture for Complex-Balanced Mass-Action Systems — FULL CLAIM
#295 Zariski Multiplicity Conjecture for Hypersurface Singularities — FULL CLAIM
#297 Amenability of Thompson's Group F — FULL CLAIM
#298 Griffiths Positivity Conjecture for Ample Vector Bundles — FULL CLAIM
#300 The Singer Conjecture on Vanishing of L2-Cohomology of Aspherical Manifolds — PARTIAL
#305 Birkhoff Conjecture (Near-Boundary Continuous Invariant Foliation) — FULL CLAIM
#306 Secret key from all entangled states — FULL CLAIM
#311 Mahler Conjecture on the Volume Product of a Convex Body — FULL CLAIM
#321 Finiteness of Finitely Presented Periodic Groups — FULL CLAIM
#328 Sinai's Positive Metric Entropy Conjecture for the Chirikov Standard Map — FULL CLAIM
#334 Unrestricted Linear-Circuit Lower Bound for the Discrete Fourier Transform — FULL CLAIM
#343 Koebe circle-domain uniformization conjecture — FULL CLAIM
#351 Kobayashi Conjecture: Hyperbolic Compact Kähler Manifolds Have Ample Canonical Bundle — FULL CLAIM
#356 Bott Conjecture: nonnegatively curved simply connected manifolds are rationally elliptic — PARTIAL
#357 Strong Hyperplane Conjecture for Isotropic Constants — FULL CLAIM
#358 CAT(0) Models for All Word-Hyperbolic Groups — FULL CLAIM
#359 Hopf Conjecture: no metric of positive sectional curvature on S^2 x S^2 — PARTIAL
#361 Alperins Weight Conjecture for Modular Representations of Finite Groups — FULL CLAIM
#364 Log-Brunn-Minkowski Inequality for Symmetric Convex Bodies — FULL CLAIM
#366 Hyperkähler SYZ (Lagrangian Fibration) Conjecture — FULL CLAIM
#368 Benjamini-Schramm conjecture p_c<p_u for non-amenable Cayley graphs — FULL CLAIM
#369 Dixmier unitarisability problem for discrete groups — FULL CLAIM
#374 Fujita Freeness Conjecture — FULL CLAIM
#375 Nagata Conjecture on Plane Curve Linear Systems — FULL CLAIM
#378 Toms–Winter: Strict Comparison Implies Z-Stability — FULL CLAIM
#379 Margulis–Platonov normal-subgroup conjecture — FULL CLAIM
#388 Shelah's Categoricity Conjecture for L_omega1,omega — PARTIAL
#392 Measure-zero singular set of stationary integral varifolds — FULL CLAIM
#401 Serres Positivity Conjecture for Intersection Multiplicities over Regular Local Rings — FULL CLAIM
#403 Rokhlin problem on multiple mixing — FULL CLAIM
#419 Lück's Determinant Conjecture for Discrete Groups — FULL CLAIM
#420 Absolute continuity of Bernoulli convolutions outside reciprocal Pisot parameters — FULL CLAIM
#421 GNRS Conjecture — PARTIAL
#424 Irrationality of Catalan's Constant — FULL CLAIM
#425 Separable Quotient Problem for Banach Spaces — FULL CLAIM
#430 Long-Time Existence of the Calabi Flow on Compact Kaehler Manifolds — FULL CLAIM
#432 Combinatorial Invariance of Kazhdan-Lusztig Polynomials — FULL CLAIM
#434 Realizable Limit Shapes in Two-Dimensional First-Passage Percolation — PARTIAL
#438 Zariski Cancellation Problem (characteristic zero, dimension ≥ 3) — FULL CLAIM
#440 Stein's Conjecture on the Hilbert Transform along Lipschitz Vector Fields — FULL CLAIM
#448 Smooth Realization of Finite-Entropy Ergodic Transformations — FULL CLAIM
#450 Deligne–Drinfeld Freeness Conjecture for grt_1 — FULL CLAIM
#453 Shub entropy conjecture — FULL CLAIM
#456 Maximal Number of Mutually Unbiased Bases in Composite Dimensions — PARTIAL
#458 LeBrun-Salamon Conjecture: positive quaternion-Kahler manifolds are symmetric spaces — FULL CLAIM
#461 Kaplansky's Direct Finiteness Conjecture — FULL CLAIM
#465 David–Semmes Riesz-transform problem in intermediate dimensions — FULL CLAIM
#467 Small Cohen-Macaulay Conjecture — FULL CLAIM
#468 Ergodicity of Irrational Triangular Billiards — FULL CLAIM
#472 CAT(0) Conjecture for All Artin Groups — FULL CLAIM
#475 Katok entropy rigidity conjecture — FULL CLAIM
#476 Polynomial-Time Solvability of Condon's Simple Stochastic Games — PARTIAL
#480 Furstenberg No-Bigeodesics (Planar FPP, Standard Minimum Moment) — FULL CLAIM
#483 Gaboriau's Fixed Price Problem for Countable Groups — FULL CLAIM
#484 Persistence Conjecture for Weakly Reversible Mass-Action Systems — FULL CLAIM
#486 Campana-Peternell Conjecture — PARTIAL
#497 Maximum Number of k-sets and Complexity of k-levels — PARTIAL
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
github.com/openai/math
YannickMrCrypto retweeted
THIS IS BIG!
The Cancer Cells That Changed Their Minds
A KAIST team just reversed cancer without killing a single cell — and the method may matter more than the result.
For a century, oncology has had exactly one strategy: find the cancer and destroy it.
Chemotherapy poisons it. Radiation burns it. Surgery cuts it out. Immunotherapy teaches the body to hunt it. Every weapon differs in precision, but not in philosophy. The tumor is an enemy. You win by killing.
Now a team at South Korea's KAIST has proposed a heresy: *what if the tumor doesn't need to die?*
In a study led by Professor Kwang-Hyun Cho of the Department of Bio and Brain Engineering, published in Advanced Science ("Control of Cellular Differentiation Trajectories for Cancer Reversion," DOI 10.1002/advs.202402132), researchers took colon cancer cells and — without poisoning, irradiating, or cutting them — turned them back into normal cells.
Not dead. Not damaged. Just normal again.
The tumors, grown in mice, shrank dramatically. The surrounding tissue was left intact, because there was never an attack to survive.
This is early research — cell lines and animal models, no human trials, real obstacles still unsolved. But the conceptual break is the story. For the first time, cancer treatment has a second verb. Not just destroy. Also: convert.
The digital twin of a cell
The KAIST team's insight begins with a redefinition of what cancer is.
The conventional view treats cancer as a pile of broken machinery: mutations accumulate, checkpoints fail, cells proliferate. Cho's group looked at the same evidence and saw something different — a trajectory. During oncogenesis, they observed, normal cells don't just break. They regress, sliding backward along the differentiation path they followed when they matured. A colon cell becomes, in effect, a confused stem cell: immature, proliferative, lost.
If cancer is a wrong turn on a developmental road, then the treatment question changes. You don't blow up the road. You build a map and find the turn.
That map is what the team calls a digital twin — a complete computational model of the gene network governing a cell's differentiation. Using data from 4,252 intestinal cells, they reconstructed a network of 522 interacting components, capturing how genes regulate one another as a cell matures or degrades.
Then they did the audacious thing: they asked the simulation which levers, flipped together, would push a cancer cell back down the road toward normalcy.
The answer came through a system they built called BENEIN (Boolean Network Inference and Control), which models gene interactions as logical relationships and systematically tests which interventions redirect the network's state. The simulation pointed to three master regulators — the genes MYB, HDAC2, and FOXA2 — acting together as the switch that holds the cancerous state in place.
Turn all three off simultaneously, the model predicted, and the cell would stop proliferating and differentiate into something resembling a normal intestinal cell.
They tested it, and it worked.
In three colon cancer cell lines, suppressing MYB, HDAC2, and FOXA2 together strongly induced differentiation into normal-like cells. The cancer cells began expressing markers of healthy intestinal tissue. Proliferation collapsed — not because the cells died, but because they grew up.
In animal models, tumors formed from the reprogrammed cells were dramatically smaller than controls, and under the microscope they looked far more like normal tissue. The signature achievement, and the one that would matter most to any patient reading this: there was no collateral damage. Nothing was poisoned, burned, or irradiated. The healthy tissue never came under fire, because there was no fire.
1 of 2
Readers added context they thought people might want to know
The KAIST 2024 study showed that inhibiting MYB, HDAC2 and FOXA2 induced differentiation in colon cancer cells to a normal-like state in cell and mouse experiments. No human trials have been done. The before-and-after scan is not from this research.
doi.org/10.1002/advs.2… eurekalert.org/news-releases/…
YannickMrCrypto retweeted
My estimates for humanoid robot production:
2027: 1 million
2028: 10 million
2029: 100 million
2030: 1 billion
2031: 10 billion
2032: 100 billion
Replying to @XRoboHub
More from Elon’s CCTV interview:
He expects at least 1 billion humanoids within 10 years, possibly sooner. His estimates then reach 10 billion in 15 years and potentially 100 billion in 20.
He explains that if robot numbers roughly double each year, that means a tenfold increase every 3–4 years.
He thinks everyone will want a personal robot: a friendly, capable helper like C-3PO or R2-D2, potentially even more capable. He mentions caring for aging parents, looking after children, one-on-one tutoring and taking on all kinds of work.
He also envisions many businesses where one person directs hundreds, thousands or even tens of thousands of physical robots and digital agents. That would vastly increase how much one person can get done.
He predicts universal high income, and a future where money itself becomes less important.
YannickMrCrypto retweeted
🔥🚨DEVELOPING: This song titled “Mace Fell Out of a Windu” has been going viral and created a discussion among entertainment fans on how ‘good AI’ is starting to become accepted after Star Wars fans instantly started sharing the track.
YannickMrCrypto retweeted
Today we’re releasing Adam and Eve, the most human-like AI voices ever built.
Adam ranks #1 among AI models on @DesignArena’s AudioRealismBench.
We believe we’ve crossed the uncanny valley. API access on our Website!
Listen: freyavoice.ai
Leaderboard: designarena.ai/leaderboard/a…
Special thanks to @alpsencerozturk, @ahmeterdempmk, and the entire Freya team for making this possible.
We’re just getting started. Join us!
YannickMrCrypto retweeted
Yesterday I said Jev would open a ton of doors...
24 hours later, this exists.
Cua built a 2.8MB model that scored 99.7% on their form-filling eval.
Hosted Jev scored 83.6%.
Not to mention it's FREE and only 706K parameters.
Small enough to run locally with not even 1gb or ram.
Fast enough to make decisions in one pass.
And specialized enough that your agent doesn't need to call a giant LLM for every tiny action.
Think about what this unlocks.
Every repetitive computer task could eventually get its own tiny specialist:
• forms
• CRM updates
• data entry
• browser actions
• document routing
• UI decisions
Then one powerful agent just routes work between them.
We are going to see some ridiculous stuff built from this as well.
1/ Introducing CUA-S1: a family of System One Models, small, specialized, and built for computer use.
Today we're open-sourcing CUA-S1-FORMS, the first in the family: github.com/trycua/cua




























