Anthropic's Claude Solves 25-Year NIST Encryption Mystery

Anthropic's AI model Claude has helped solve a 25-year-old cryptography mystery related to NIST encryption standards. The model identified that suspicious seed values came from simple Unix timestamps fed into weak 1990s random number generators. The discovery confirmed no backdoor existed, reassuring users who long suspected hidden manipulation in cryptographic standards.

Claude's research also sheds light on why Bitcoin uses the secp256k1 curve while devices like iPhones rely on P-256. Since the NIST seeds were generated from basic timestamps, Satoshi Nakamoto likely chose a different curve for Bitcoin out of distrust. New tools like RIP-7212 now help bridge this gap, with companies like Optimism and Arbitrum supporting the effort.

In another development, AI analysis of federal court decisions reveals significant inconsistencies in how judges apply early release standards. Over 100,000 people are currently on federal supervised release, and the legal criterion depends on the vague phrase "interest of justice." One AI agent processed individual decisions in about five seconds, exposing wide variations between districts such as Virginia and Utah.

AI coding teams are adopting new methods to improve reliability. About 46% of developers distrust AI tools due to inaccurate output. Teams now use path-scoped rules to guide agents on legacy code, while LangGraph workflows add determinism by running steps in a fixed order. This approach allows teams to handle large tasks in under a day.

Autonomous AI newsrooms are transitioning from experiments to real publications, with roughly 30 such outlets worldwide. RuntimeWire uses AI to beat competitors on breaking news, while Redazione Zero in Italy operates multi-agent workflows with almost no human input. Moj Kraj in Croatia represents another practical example of AI-driven journalism.

Concerns about AI being weaponized for control are growing. Some leaders use alarming AI warnings to push rules that disadvantage open-source AI, helping big tech maintain power while developing nations remain dependent on costly tools. Open-source AI supports real needs in healthcare, farming, and energy, while focus on imaginary dangers distracts from problems AI already causes in jobs and communication.

Tweedy Risk Information Consulting has reopened after a two-year pause. President David Tweedy brings nearly 50 years of experience in risk technology, helping companies evaluate software for workers' compensation, liability, and property risks. He notes that many insurance companies discuss AI but have not yet seen tangible results from its adoption.

Humanoid robots powered by AI are advancing rapidly, with Boston Dynamics' Atlas as a prominent example. Hyundai has invested in Boston Dynamics, and Atlas learns human movements through training while using NVIDIA chips to process information. The robot Spot already assists in factory inspections and quality checks. Experts believe humanoid robots will assist people rather than replace them.

The German Bishops' Conference has joined the KI kindgerecht initiative to make AI safer for children. The effort follows Pope Leo XIV's call to protect human dignity in the age of AI. Germany's largest Protestant body and children's website groups are also involved. Bishop Georg Bätzing emphasized that AI is already part of many digital tools, and the initiative aims to ensure development considers children's needs.

OpenAI has hired Graham Seamans, who led xAI's sales team since August 2025, to join its cybersecurity business. The move supports OpenAI's expansion of enterprise sales and exploration of new revenue streams as the company prepares for a potential public listing. Seamans' departure from xAI comes as the startup faces growing competition from other AI companies.

Key Takeaways

Claude solved a 25-year-old cryptography mystery by identifying that NIST seed values came from simple Unix timestamps and weak 1990s random number generators, confirming no backdoor existed.
Claude's research explains why Bitcoin uses secp256k1 while iPhones use P-256, with Satoshi Nakamoto likely choosing a different curve due to distrust of NIST standards.
AI analysis of over 100,000 federal supervised release cases reveals huge inconsistencies in how judges apply the "interest of justice" standard across districts like Virginia and Utah.
About 46% of developers distrust AI coding tools, but path-scoped rules and LangGraph workflows now help teams handle large tasks in under a day with improved reliability.
Roughly 30 autonomous AI newsrooms now operate worldwide, including RuntimeWire, Redazione Zero in Italy, and Moj Kraj in Croatia, shifting from experiments to real publications.
Some leaders use AI fear to push rules that hurt open-source AI, helping big tech maintain power while developing nations stay dependent on expensive tools.
Tweedy Risk Information Consulting reopened after a two-year break, with President David Tweedy bringing nearly 50 years of risk technology experience.
Boston Dynamics' Atlas humanoid robot uses NVIDIA chips and learns human movements through training, while Spot already helps inspect factories and perform quality checks.
The German Bishops' Conference joined the KI kindgerecht initiative to make AI safer for children, following Pope Leo XIV's call to protect human dignity.
OpenAI hired Graham Seamans, former xAI sales lead, to strengthen its cybersecurity business and enterprise sales ahead of a potential public listing.

Claude solves 25-year-old encryption mystery from clock data

Anthropic says its AI model Claude solved a 25-year-old cryptography mystery about NIST encryption standards. The suspicious seed values came from a simple Unix timestamp fed into a weak 1990s random number generator. Claude found the origin by scanning patterns and using math techniques. The discovery reassures users because no backdoor was found.

Claude finding explains why Bitcoin and iPhone use different curves

Claude's research explains why Bitcoin uses secp256k1 while devices like iPhones use P-256. The NIST seeds were generated from basic timestamps, not hidden weaknesses. This history of distrust likely pushed Satoshi Nakamoto to choose a different curve for Bitcoin. New tools like RIP-7212 now help bridge this gap for modern wallets. Companies like Optimism and Arbitrum support this bridge too.

AI reveals inconsistent logic in federal early release decisions

AI analysis shows federal judges apply the early release standard very differently. Over 100,000 people are on federal supervised release. The legal standard relies on the vague phrase interest of justice. One AI agent processed individual court decisions in about five seconds each. The data reveals huge differences between districts like Virginia and Utah.

AI coding teams use path rules to fix code quality

A new approach brings reliability back to AI-assisted coding. 46% of developers distrust AI tools due to inaccurate output. Teams now use path-scoped rules to guide agents on legacy code correctly. LangGraph workflows add determinism by running steps in a fixed order every time. This setup allows teams to handle large tasks in under a day.

AI-native newsrooms shift from experiments to real publications

A growing number of newsrooms now use autonomous AI to find and publish stories. About 30 such outlets exist worldwide. RuntimeWire uses AI to beat competitors on breaking news. Redazione Zero in Italy uses multi-agent workflows with almost no human input. Moj Kraj in Croatia is another example of practical AI news use.

AI fear is being used to control the world

Some leaders and companies use scary AI warnings to push for rules that hurt open-source AI. This helps big tech companies keep power while developing nations stay dependent on expensive AI tools. Open-source AI is important for countries that cannot afford to build their own systems. It can help with real needs like healthcare, farming, and energy. Focusing on imaginary dangers takes attention away from real problems AI already causes in jobs, the environment, and communication.

Tweedy Risk Consulting returns after two year break

Tweedy Risk Information Consulting has reopened after a two year pause. President David Tweedy brings nearly 50 years of experience in risk technology. He helps companies evaluate software for managing workers' compensation, liability, and property risks. He also provides claims auditing and expert advice. He believes many insurance companies talk about AI but have not seen real results yet.

Letter questions AI race with China

A letter to the editor questions the US claim that it is racing China in AI. The writer asks why China is not building data centers if it is truly competing. The short letter raises doubts about how the AI race is being measured. It suggests the real focus of the competition may be misunderstood.

Humanoid robots are helpers not enemies

Engineers are advancing humanoid robots powered by AI, with Boston Dynamics' Atlas as a key example. Hyundai has invested in Boston Dynamics to help develop these robots. Atlas learns human movements through training and uses NVIDIA chips to understand the world. The robot Spot already helps inspect factories and perform quality checks. Despite fears, experts believe humanoid robots will assist people rather than replace them.

German bishops back child-friendly AI plan

The German Bishops' Conference joined an initiative called KI kindgerecht to make AI safer for children. The effort follows Pope Leo XIV's call to protect human dignity in the age of AI. Germany's largest Protestant body and children's website groups are also involved. Bishop Georg Bätzing said AI is already part of many digital tools. The initiative aims to ensure AI development considers children's needs.

OpenAI hires xAI sales lead for cybersecurity team

OpenAI hired Graham Seamans, who led xAI's sales team since August 2025, to join its cybersecurity business. The move supports OpenAI's expansion of enterprise sales and exploration of new revenue streams. The company is preparing for a potential public listing. Seamans' departure from xAI comes as the company faces growing competition from other AI startups.