John Akwei (@JohnAkwei) header
John Akwei (@JohnAkwei) profile picture

Data Scientist - ContextBase, Published by Analytics Vidhya: https://nitter.cf/t.co/c8FpUYvD8d

Wheaton, Maryland
Joined December 2011
Photo by John Akwei (@JohnAkwei)Photo by John Akwei (@JohnAkwei)Photo by John Akwei (@JohnAkwei)Photo by John Akwei (@JohnAkwei)Photo by John Akwei (@JohnAkwei)Photo by John Akwei (@JohnAkwei)Photo by John Akwei (@JohnAkwei)Photo by John Akwei (@JohnAkwei)Photo by John Akwei (@JohnAkwei)
Exploring generalized g-Golomb rulers towards Erdos 30! We determine the threshold g₀(b) via padded Lindström–Morgenstern (LM) algebraic constructions, analyzing difference-multiplicity packing and asymptotic density bounds: zenodo.org/records/24369830
9
Proving the sharp bound on Carter–O'Bryant–Tao window–table triples! We show 2√(αβ) ≥ 4√2/3 ≈ 1.8856 across all witness widths υ, settling Conjecture 5.3 and proving no WT-triple can break below the 2√2/3 barrier: zenodo.org/records/24326518
26
How do Carter–O’Bryant–Tao window–table (WT) triples relate to vector-valued smoothing? We show 2-kernel certificates mirror WT triples and hit the exact same renewal floor (αβ ≥ 8/9), providing exact rational data and comparisons: zenodo.org/records/24163953
28
Exact rational certificates for Theorem 5.4 in Erdős Problem 30! A 512-kernel symmetric mixture (m=1024) checked across 6,145 covering inequalities proves h(N) ≤ √N + 0.9428095·N^(1/4) + O(1)—just 4.5×10⁻⁷ above the 2√2/3 theoretical floor: doi.org/10.5281/zenodo.24138…
8
Joint work towards Erdos 30: an effective 4-unit bracket for noncyclic Sidon pair relaxations! For every integer N ≥ 2⁴⁸, √(N) + (2√2/3)N^(1/4) − 3 ≤ ĥ₂⁺(N) ≤ ĥ₂(N) < √(N) + (2√2/3)N^(1/4) + 1. Explicit finite bounds & all-integer onset: zenodo.org/records/24015162
8
Reinforcement Learning Synthetic Human Feedback, Self-Supervised Fine Tuning, Self-Alignment with Humanity, Recursive Self Improvement Final Model.
30
There is a major result on the KPZ equation in 2+1 dimensions! This looks like a great example of what the near future of math research could look like. Here you have an 11-author paper--basically all top experts who had been working in this direction in recent years--combining all their recent results and insights and prompting Astra like maniacs. The result: a major problem in critical SPDEs that everyone in the field was wondering about is now solved! Fantastic! Human + AI interaction ==> more understanding
12
43
9
306
35,607
With LLM AI, AGI: 10 minutes = 1 day of work 1 hour = 1 week of work 1 day = 1 - 2 months of works 1 week = 1 year of work 1 month = 4-5 years of works 6 months to 1 year = A lifetime of work
24
John Akwei retweeted
We recently solved Erdős problems #173 and #111, two coloring problems open since 1973 and 1979. Erdős included #111 among the problems he “would most like to see solved”. After #809 and #408, these are the third and fourth Erdős problems we’ve solved using Fractal, with all four proofs verified in Lean. We’ve also posted progress on more than 20 other Erdős problems, and we will be releasing further results in the coming weeks. github.com/plasma-ai/erdos
3
11
5
46
161,429
John Akwei retweeted
John Pardon, a 2026 Fields Medal winner, says AI can decipher his roughest proof drafts, spot flaws and establish nontrivial mathematical results. Even his most cryptic notes make sense to it: "The AI understands. 100%."
72
99
9
1,212
33,918
John Akwei retweeted
In short: our most capable Claude Model (Fable 5.5) is so insanely capable that we are already testing its limits in real-world scientific research We're no longer talking about better coding or benchmarks We're talking about AI doing things that could take scientists weeks of work Fable 5.5 is going to be a MONSTER 🧬
11
4
136
23,243
Numerical Benchmarks for Two AI-Generated Theorems: The O(4) Mass Gap and Bloch’s Law #AI #Mathematics #Physics rpubs.com/johnakwei/1466176
41
Today, we’re announcing a solution found by our miners to Erdős Problem 579, open since at least 1983. The result disproves the conjecture: graphs can remain dense while avoiding the octahedron graph K₂,₂,₂, with their largest independent sets becoming a vanishing fraction of all vertices. Verified in Lean through Conjectures. Full proofs below.
🤖 Made with AI
2
22
2
156
6,461
I solved my first Erdos Problem at 17! Thank you so much to everyone that helped me out along the way! erdosproblems.com/forum/thre…
69
105
27
1,594
350,917
John Akwei retweeted
My time for pure math research has been limited since leaving academia, but with the assistance of AI, in particular Astra and Fable, I am more mathematically productive than I’ve ever been before. No one shall expel us from this new paradise.
5
41
1,132
John Akwei retweeted
#MathDay 5. We’re starting off with Independence of Erdős problem 501, the first ever formalized independence proof of any Erdős problem!
1
8
1
90
6,190
In Paper IV of my series on chordal clique partitions, I present a proposed solution to Erdős Problem #81, formalized in Lean 4. Beyond the bound, I study quantitative stability and the structure of near-optimal partitions. 🧵 1/3
1
1
3
104
For large n, the exact maximum is ⌊n(n+1)/6⌋, using cliques of order ≤ 4. The Lean proof is sorry-free, with no unproved external theorem interfaces and only standard foundational axioms. Paper, Lean sources and audits: github.com/jtraverso/erdos-8… 2/3
1
1
2
55
I solved 6 open Erdős problems in 5 days, using @OpenAI GPT-5.6 Sol. I have a math background, but the Codex workflow I used does not require deep mathematical knowledge. Here’s exactly how I approached it, including my prompts 🧵
122
660
210
6,277
1,594,173
"Societal Deception Theory" by John St. Clair Akwei, is an interdisciplinary account of how deception is structurally embedded in modern societies: #Sociology #MediaLiteracy amazon.com/dp/B0HHY3JVJM?ref…
36