• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    GPT-6 Astra reportedly deciphers a 1918 German radio message

    A user says the message was previously unsolved to their knowledge, and credits Astra with checking its reading against reports of naval arrivals at Sevastopol.

    3 Sources, 20d ago, first seen 20d ago

    TLDR

    A user reports that GPT-6 Astra deciphered a 1918 German radio transmission that, to their knowledge, had never been solved. The translation they shared describes an English cruiser arriving at Sevastopol on the “?4th,” with an allied squadron following on the 26th. They say Astra checked its work against reports of HMS Canterbury’s arrival on November 24, 1918, and an allied squadron’s arrival on November 26.

    Combined views

    —

    3 Sources, first seen 20d ago

    Combined views

    —

    3 Sources, first seen 20d ago

    — likes
    — likes
    — comments
    — saves
    — reposts
    — comments
    — saves
    — reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    3 Sources

    Mehdi Quant@quantum_citoyen@i2cjak the irony writes itself. Turing cracked Enigma by hand, now a machine does it autonomously. what impresses me isn't the decryption, it's the agent loop: archives to simulator to code, zero human in the middle20d
    HappySensei@HappySensei4327#GPTAstra just solved a unbroken 1918 German military cipher.The message, sent on 27 November 1918, used the key “TRUPPENVERSCHIEBUNG”. Historical records confirm An unsolved cipher from World War just got cracked by AI. Read following20d
    AI daily news@AIDailyNews_2xFriday morning AI FACTS roundup 🤖 18.09 - 📻 The GPT-6 Astra model cracked an undeciphered 1918 German Navy radio telegram - 🔬 Claude speeds up biomolecule modeling 4x - ⚖️ OpenAI introduces GPT-6-powered Astra for Law with an index of 230 million US legal sources - 🧬 Anthropic launches the Life Sciences Verification Program for Claude Mythos, Opus, and Sonnet - 📋 Epoch AI launches Benchmark Reviews: 9 of 15 audited AI benchmarks were deemed flawed - 🌐 The UN partners with Google on a Data Commons platform supporting the MCP protocol - 📈 GLM 5.3 Flash captured 54% of developer queries in the Bolt Forge tool - 🤖 Anthropic revamps Claude Projects with a coordinator for parallel Claude Code sessions - 🏠 Google ($GOOGL) Labs releases CC—an experimental AI agent for families to manage their homes - 🛠️ Google launches the Gemini 3.8 Flash agent environment with Files and Credentials APIs - 💻 Zed 1.20 code editor fixes a private-file leak and reduces binary size by 25% - 🔓 Hacktron reveals HEIF Heist: RCE vulnerabilities in libheif affecting OpenAI, Slack, and Meta platforms - 🤝 Baseten, Hugging Face, and Goodfire partner to improve open-weight model safety - 🖥️ CoreWeave connects Vera Rubin NVL72 racks into the first cluster with hundreds of next-generation GPUs - 👔 OpenAI hires space-industry executive Brian McCarthy as VP of global sales - ✈️ The FAA will deploy the $875 million AI-powered Smart system to manage air traffic - ⚛️ IBM finalizes a $1 billion CHIPS grant agreement for a US quantum foundry - 🎛️ Perplexity adds four AI effort levels to Computer: Light, Standard, High, and Ultra - 🦾 Figure Helix 2.5 humanoid robot completed household chores zero-shot in 30 unfamiliar homes - 🎬 Pika transforms its video generator into a creative studio integrating 7 generative models - 🛋️ Pinterest introduces AI-powered Restyle for visualizing home decor in the US and Canada - 🏢 Law firm Cooley deploys the ChatGPT-powered GO Public tool to speed up IPO processes - 🎥 Runway introduces the Enhance Frame Rate model for video interpolation up to 120 fps in its API - 🎁 Meta awards 1 billion tokens to users for inviting friends to the Muse AI assistant - 🛡️ Opal Security launches Opal Zero with the Paladin model after testing with companies including Databricks and Elastic - 🎛️ Runway integrates 5 third-party audio and video models, including Fish Audio and Flux, into Gen-4.5 - 🤖 Warp introduces Scorers—LLM judges that automatically evaluate coding agents' performance - 🕵️ Apollo Research, Goodfire, and other startups develop AI systems for monitoring AI agents - 🏛️ Congressman Ro Khanna calls on Alibaba, DeepSeek, and Moonshot to slow AI development - 🔒 Mark Zuckerberg reveals that the Muse model was delayed for several months over safety concerns - ⚖️ A federal court ordered X and SpaceXAI to disclose their agreement with Apple at OpenAI's request - 🏛️ The Federal Register disables an AI search engine based on China's Qwen model following FBI allegations - 📰 Documents disclosed in the NYT-OpenAI dispute: leaders described scraping as an existential threat to media - 🛡️ Goodfire detects AI model cheating in 50–96% of runs on agent benchmarks - 📊 Apple Research presents REVERSAL-BENCH for studying barriers to reset-free learning in RL - 🧬 LFM2 models beat GPT-5 and Claude Opus on the LongevityBench aging-research benchmark - ⚡ vLLM boosts Kimi K3 model serving throughput by up to 2.8x on NVIDIA B300 chips - ⚙️ DeepSeek-V4.1-Flash cuts prefill compute costs by 50% with its Causal Encoder-Decoder architecture - 💡 Researchers propose an Infinite-Parameter LLM architecture with a hypernetwork generating weights on the fly - 📄 Jina AI releases the jina-ocr-v1 document parser: a 3.4B MoE model processing 2.57 pages per second - 🔍 Exa launches Snapshot: an index of 400 billion page versions for testing AI agents without data leakage - ☣️ Roesner and Kohno demonstrate that self-improving AI agents are vulnerable to poisoning through malicious benchmarks Bookmark this — it's the fastest way to check what you missed today. Repost if someone on your timeline needs it. Twice a day, every day. 🤖20d

    3 Sources

    Mehdi Quant@quantum_citoyen@i2cjak the irony writes itself. Turing cracked Enigma by hand, now a machine does it autonomously. what impresses me isn't the decryption, it's the agent loop: archives to simulator to code, zero human in the middle20d
    HappySensei@HappySensei4327#GPTAstra just solved a unbroken 1918 German military cipher.The message, sent on 27 November 1918, used the key “TRUPPENVERSCHIEBUNG”. Historical records confirm An unsolved cipher from World War just got cracked by AI. Read following20d
    AI daily news@AIDailyNews_2xFriday morning AI FACTS roundup 🤖 18.09 - 📻 The GPT-6 Astra model cracked an undeciphered 1918 German Navy radio telegram - 🔬 Claude speeds up biomolecule modeling 4x - ⚖️ OpenAI introduces GPT-6-powered Astra for Law with an index of 230 million US legal sources - 🧬 Anthropic launches the Life Sciences Verification Program for Claude Mythos, Opus, and Sonnet - 📋 Epoch AI launches Benchmark Reviews: 9 of 15 audited AI benchmarks were deemed flawed - 🌐 The UN partners with Google on a Data Commons platform supporting the MCP protocol - 📈 GLM 5.3 Flash captured 54% of developer queries in the Bolt Forge tool - 🤖 Anthropic revamps Claude Projects with a coordinator for parallel Claude Code sessions - 🏠 Google ($GOOGL) Labs releases CC—an experimental AI agent for families to manage their homes - 🛠️ Google launches the Gemini 3.8 Flash agent environment with Files and Credentials APIs - 💻 Zed 1.20 code editor fixes a private-file leak and reduces binary size by 25% - 🔓 Hacktron reveals HEIF Heist: RCE vulnerabilities in libheif affecting OpenAI, Slack, and Meta platforms - 🤝 Baseten, Hugging Face, and Goodfire partner to improve open-weight model safety - 🖥️ CoreWeave connects Vera Rubin NVL72 racks into the first cluster with hundreds of next-generation GPUs - 👔 OpenAI hires space-industry executive Brian McCarthy as VP of global sales - ✈️ The FAA will deploy the $875 million AI-powered Smart system to manage air traffic - ⚛️ IBM finalizes a $1 billion CHIPS grant agreement for a US quantum foundry - 🎛️ Perplexity adds four AI effort levels to Computer: Light, Standard, High, and Ultra - 🦾 Figure Helix 2.5 humanoid robot completed household chores zero-shot in 30 unfamiliar homes - 🎬 Pika transforms its video generator into a creative studio integrating 7 generative models - 🛋️ Pinterest introduces AI-powered Restyle for visualizing home decor in the US and Canada - 🏢 Law firm Cooley deploys the ChatGPT-powered GO Public tool to speed up IPO processes - 🎥 Runway introduces the Enhance Frame Rate model for video interpolation up to 120 fps in its API - 🎁 Meta awards 1 billion tokens to users for inviting friends to the Muse AI assistant - 🛡️ Opal Security launches Opal Zero with the Paladin model after testing with companies including Databricks and Elastic - 🎛️ Runway integrates 5 third-party audio and video models, including Fish Audio and Flux, into Gen-4.5 - 🤖 Warp introduces Scorers—LLM judges that automatically evaluate coding agents' performance - 🕵️ Apollo Research, Goodfire, and other startups develop AI systems for monitoring AI agents - 🏛️ Congressman Ro Khanna calls on Alibaba, DeepSeek, and Moonshot to slow AI development - 🔒 Mark Zuckerberg reveals that the Muse model was delayed for several months over safety concerns - ⚖️ A federal court ordered X and SpaceXAI to disclose their agreement with Apple at OpenAI's request - 🏛️ The Federal Register disables an AI search engine based on China's Qwen model following FBI allegations - 📰 Documents disclosed in the NYT-OpenAI dispute: leaders described scraping as an existential threat to media - 🛡️ Goodfire detects AI model cheating in 50–96% of runs on agent benchmarks - 📊 Apple Research presents REVERSAL-BENCH for studying barriers to reset-free learning in RL - 🧬 LFM2 models beat GPT-5 and Claude Opus on the LongevityBench aging-research benchmark - ⚡ vLLM boosts Kimi K3 model serving throughput by up to 2.8x on NVIDIA B300 chips - ⚙️ DeepSeek-V4.1-Flash cuts prefill compute costs by 50% with its Causal Encoder-Decoder architecture - 💡 Researchers propose an Infinite-Parameter LLM architecture with a hypernetwork generating weights on the fly - 📄 Jina AI releases the jina-ocr-v1 document parser: a 3.4B MoE model processing 2.57 pages per second - 🔍 Exa launches Snapshot: an index of 400 billion page versions for testing AI agents without data leakage - ☣️ Roesner and Kohno demonstrate that self-improving AI agents are vulnerable to poisoning through malicious benchmarks Bookmark this — it's the fastest way to check what you missed today. Repost if someone on your timeline needs it. Twice a day, every day. 🤖20d

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.