This article reports on Donald Trump's pressure on Ukraine to halt attacks on Russian oil facilities. This move is linked to Washington's intention to curb US fuel prices and stabilize energy markets ahead of the November congressional elections.

However, this decision complicates Ukraine's defense strategy, especially coinciding with restrictions on Starshield (SpaceX's military satellite communication service) and a shortage of Patriot missiles. Ukraine views attacks on Russian oil infrastructure as crucial to pressure Russia's war efforts but faces a dilemma due to its reliance on US support.

This article has no direct relevance to topics such as Generative AI, Image Generation AI, Multimodal AI, AI Agents, LLM, or Vibe Coding.

👉 Read the full article on Gizmodo en Español

Research by Anthropic reveals that its LLM, Claude, has solved a 25-year-old mystery regarding the origins of ECDSA (Elliptic Curve Digital Signature Algorithm) curve seeds. These seeds, adopted by NIST in 2000 as part of the FIPS 186-2 Digital Signature Standard, had been a source of long-standing concern and speculation due to their unexplained provenance from the NSA.

Claude discovered that the seeds were not generated by a hash algorithm, as previously rumored, but rather by the C library's lrand48 pseudorandom number generator, seeded with a Unix timestamp from 1997-1998. This finding increases confidence that the curves were chosen without backdoors or intentional weaknesses, closing a significant chapter for the cryptographic community.

Furthermore, Claude also identified the origins of some base points G, tracing them to random number generators in FreeLIP and NTL libraries. This achievement clearly demonstrates AI's capability to solve complex mathematical and cryptographic problems, accelerating scientific discovery.

👉 Read the full article on Anthropic

Research by Anthropic demonstrates that its LLM, Claude, has successfully designed more efficient quantum circuits. Given plain-language specifications for eight circuits from fields like chemistry, optimization, physics simulation, and number theory, along with an automated evaluator, Claude found designs that significantly improved upon the best previously published circuits.

This AI agent operated in a closed loop, proposing circuit designs, receiving feedback and scores from the evaluator, and using that information to refine its subsequent attempts. Notably, Claude uncovered new structural and mathematical insights in areas such as modular inversion and the most expensive steps of quantum optimization algorithms, substantially reducing the required number of qubits and gates.

This achievement suggests that AI can accelerate advancements in quantum computing, potentially bringing practical quantum applications, such as designing new medicines and materials, within reach sooner than previously anticipated. Claude's capabilities prove its value in programming and optimization under complex constraints.

👉 Read the full article on Anthropic