cd /news/artificial-intelligence/asked-ai-to-do-a-mini-math-breakthro… · home topics artificial-intelligence article
[ARTICLE · art-92711] src=github.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

Asked AI to do a mini math breakthrough. It did a mini breakthrough

A second AI system has produced a candidate improvement to the 67.25% zeta-zero bound, refining Anthropic's Theorem D. The new repository packages a stronger argument with an exact interval-arithmetic verifier, and a clean-room reproduction of the computer-assisted lemma. The candidate bound is unreviewed and does not prove the Riemann hypothesis.

read2 min views1 publishedAug 11, 2026
Asked AI to do a mini math breakthrough. It did a mini breakthrough
Image: source

A reproducible candidate improvement to the 67.25% zeta-zero bound.

One AI system produced Anthropic's new Theorem D. A second AI-generated research draft found a small strengthening. This repository packages the stronger argument, its exact interval-arithmetic verifier, and a clean-room reproduction of the only computer-assisted lemma.

The candidate bound is

Here

Important

This is an unreviewed candidate refinement. It does not prove the Riemann hypothesis and it does not independently replace the analytic results imported from Anthropic's Theorem D. The finite interval certificate has been reproduced exactly; broader expert review is still needed.

Paper (PDF) · LaTeX source · technical audit · reproduction guide

The strengthened source is isolated in commit 0faebf367ffb780951845d014e2e3d9a4a04adb1. Starting from the public verifier, it changes the target to

A clean x86_64 Linux run with CPython 3.12.3 and python-flint==0.8.0

returned:

verified=true
target=F6 >= 191/50000
grid=4000
precision_bits=128
initial_boxes=729
nodes=786215
pruned=393472
splits=392743
maximum_depth=43
kernel_table_sha256=f79a9147ffc37692b84330e98422fb94cbf601c51d4f9f8ae749982f23838571
second_derivative_table_sha256=cc98102590dba6e1a982a5a4c9fcd755848a93483db67edd6634813a35e5e3f5

The exact expected report is committed in certificates/seven-point.expected.json. CI reruns the exhaustive verifier from a clean checkout and compares every deterministic field.

Anthropic's Theorem D gives

The seven-point refinement proves, using those imported analytic inputs,

for all nonnegative gaps. With block size

The matrix inequality, kernel normalization, seven-point combinatorics, shifted-block pinching, and final arithmetic are covered in the audit. The analytic trace estimates, tail bounds, and optimized test family remain dependencies of the cited Anthropic paper and its Lean 4 artifact.

The most exact route uses the pinned container image:

docker build -t ai-refines-ai-zeta-bound .
docker run --rm ai-refines-ai-zeta-bound

For a local CPython 3.12 installation:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --require-hashes -r requirements.lock
PYTHONPATH=src python -m unittest discover -s tests -v
PYTHONPATH=src python scripts/verify_release.py

The exhaustive verification usually takes a few minutes. An unresolved terminal cell is a hard failure; the program never treats it as verified.

  • Upstream artifact: , commitainta/zeta-simple-zeros

040c5e899e658aed7b56a2a87f501798fe10761d

. - The upstream repository identifies the research draft as generated by GPT-5.6 Sol.

AI-generated research draft. Computer-assisted lemma independently reproduced. No human peer review is claimed. Attempts to falsify the argument or reproduce the certificate on another architecture are welcome once the repository is made public.

MIT, inherited from the upstream artifact.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @anthropic 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/asked-ai-to-do-a-min…] indexed:0 read:2min 2026-08-11 ·