Hasty Briefsbeta

Bilingual

Let Us Learn to Show Friendship for a Man When He Is Alive, Not After He Is Dead
5 hours ago
- Sam Altman is the subject of a lengthy New Yorker profile questioning his trustworthiness, particularly regarding his 2023 ouster and return to OpenAI's board.
- The article reports that close colleagues, including Y Combinator partners and Microsoft executives, express deep mistrust, with some describing Altman as 'sociopathic' and comparing him to Bernie Madoff or Sam Bankman-Fried.
- Aaron Swartz, a former Y Combinator cohort member, reportedly warned friends that Altman is a 'sociopath' who 'can never be trusted,' adding significant weight to concerns about his integrity.
- OpenAI's leadership structure is scrutinized, with Fidji Simo's role as 'CEO of AGI Deployment' seen as potentially meaningless, and her recent medical leave sparking theories of internal power struggles.
- Paul Graham publicly denies firing Altman from Y Combinator, emphasizing a choice between roles, but avoids directly affirming Altman's honesty or integrity, leaving trust questions unanswered.
- The piece suggests Altman's financial structuring at OpenAI may be alarmingly circular, drawing parallels to Enron, despite the real technological advancements like ChatGPT and Codex.
FBI Extracts Suspect's Deleted Signal Messages Saved in iPhone Notification Data
5 hours ago
- FBI extracted deleted Signal messages from an iPhone's push notification database after the app was deleted.
- The case involved a vandalism incident at an ICE facility and an officer shooting in Texas, with forensic extraction revealing sensitive data.
- Signal has a setting to hide message content in notifications; enabling it could prevent such data storage.
- Only incoming messages were recovered from the notification storage, not outgoing ones, even for messages set to disappear in Signal.
- Discussions highlight user responsibility in managing notification settings and potential design flaws in data retention.
Social media has become a freak show
4 hours ago
- Author announces a Substack Live session on Trump/Iran policy and opens questions for SBSQ #31.
- Recalls FiveThirtyEight's launch under Disney/ESPN in 2014, a time when quality content was undervalued due to the focus on maximizing reach via Facebook.
- Critiques Facebook's algorithm for rewarding emotional headlines and notes that FiveThirtyEight's analytical style didn't align, making viral traffic low-quality.
- Describes Twitter in the mid-2010s as better for FiveThirtyEight, rewarding expertise and nerdy argumentativeness, but later became more partisan and woke.
- Notes social media's declining relevance for media strategy, with Silver Bulletin's external social traffic at 0.7%, yet overall views increased significantly.
- Highlights that current Twitter engagement is dominated by low-quality, partisan accounts like Catturd and Elon Musk, with even liberal accounts being combative.
- Compares social media ecosystems to ecology, where algorithms shape content survival, and notes current platforms like X and Bluesky are unhealthy or siloed.
- Mentions the 'island effect' in ecology to explain Twitter's strange, low-quality influencers thriving in an isolated, non-competitive environment.
- Concludes by affirming Silver Bulletin as an 'island of sanity' and plans to return to topics like Trump and Iran.
PGLite Evangelism
2 days ago
- The text appears to describe an app or website interface elements, including options like Home, Subscriptions, Chat, Activity, Explore, Profile, and Create.
- The app is described as 'for independent voices' and prompts users to 'Get started' or 'Learn more'.
- A section labeled 'For you' suggests personalized content, and the mention of 'Get app' indicates availability as a mobile application.
- A technical note indicates that the site requires JavaScript to function properly and asks users to enable JavaScript or unblock scripts.
Artemis II safely splashes down
4 hours ago
- Artemis II crew returned to Earth with a splashdown in the Pacific Ocean off San Diego after a 9-day moon mission.
- The Orion capsule, carrying astronauts Reid Wiseman, Victor Glover, Christina Koch, and Jeremy Hansen, set a record for the farthest human distance from Earth at 252,756 miles.
- Reentry involved a communications blackout due to plasma, peak temperatures of 5,000°F on the heat shield, and parachute deployment for a gentle splashdown at 8:07 p.m. ET.
- NASA and military recovery teams approached Orion post-splashdown for crew extraction, medical checks on USS John P. Murtha, and capsule retrieval.
- President Trump watched the splashdown and praised the astronauts as modern-day pioneers, highlighting plans for future moon and Mars missions.
- The mission included emotional moments, such as naming a moon crater after Commander Wiseman's late wife and capturing photos of an eclipse and Earth from space.
Show HN: See what your employees are prompting LLMs (without network proxies)
a day ago
- Privent provides a security layer embedded within AI agent execution graphs (e.g., LangGraph, CrewAI, n8n) to protect data before reaching external models.
- It intercepts and controls data movements by reading full runtime states, including message history, tool outputs, and context, ensuring sensitive information doesn't leave infrastructure.
- Instead of blocking data, it transforms sensitive payloads using ACARS scoring and APE techniques like tokenization or substitution, maintaining pipeline functionality.
- Deployment is lightweight and non-disruptive, offering a 30-day silent assessment to map data risks across platforms like ChatGPT, Claude, and Gemini.
- Privent logs detection events with risk signals and compliance data (GDPR, HIPAA, EU AI Act) without storing raw prompts, facilitating audits.
- It secures all points where enterprise data meets AI, from employee prompts to agent tool calls, preventing leaks across agent boundaries.
- Pricing is custom, tailored to deployment size and compliance needs, starting with a free Enterprise AI Risk Report.
Sam Altman's response to Molotov cocktail incident
5 hours ago
- A Molotov cocktail was thrown at the author's house at 3:45 am, but no one was hurt.
- The author reflects on underestimating the power of words and narratives, prompted by an incendiary article about them.
- Beliefs include: working towards prosperity for all and advancing science as moral obligations; AI as a powerful, uncapped-demand tool; justified fears about AI requiring societal resilience; democratizing AI to avoid concentration of power; and the need for adaptability in beliefs as technology evolves.
- Personal reflections: proud of resisting Elon Musk's unilateral control over OpenAI and navigating the company's survival, but not proud of being conflict-averse and mishandling conflicts with the board, apologizing for mistakes and harm caused.
- OpenAI has transitioned from a startup to a major platform, achieving its mission against odds by building powerful AI, securing capital, and scaling safe services.
- Industry thoughts: AGI creates a 'ring of power' dynamic, leading to drama; solution involves broadly sharing technology and ensuring democratic control, with laws and norms adapting through a messy process.
- Criticism of the AI industry is valid due to high stakes, but the author advocates for de-escalating rhetoric and tactics to reduce conflicts, both figuratively and literally.
A Cancer Treatment That Does More Than Scientists Thought
5 hours ago
- CAR-T cell therapy, originally developed for cancer, is showing promise for treating autoimmune diseases.
- A 47-year-old patient with three severe autoimmune diseases achieved remission after CAR-T treatment, returning to a mostly normal life.
- CAR-T therapy involves reengineering T cells to target and eliminate misbehaving B cells, potentially resetting the immune system.
- Early studies and trials indicate CAR-T can induce remission in various autoimmune diseases like lupus, with few side effects.
- Challenges include high costs, limited access in some countries, and potential risks such as immunocompromised states.
- Researchers are exploring more efficient and affordable methods, including vaccine-like injections to generate CAR-T cells.
- Long-term outcomes are still uncertain, but some patients remain in remission years after treatment.
Installing Every* Firefox Extension
6 hours ago
- Successfully scraped and attempted to install every Firefox extension, revealing a total of approximately 84,194 extensions.
- Identified the largest extension at 196.3 MB, while the smallest was just 7,518 bytes with no code.
- Discovered significant issues including phishing extensions, SEO spam, and potentially unwanted applications (PUAs) with over 700,000 total users.
- Encountered performance challenges when installing all extensions, with Firefox becoming unstable beyond 6,000 extensions and requiring extensive memory (27-37 GiB).
- Noted that 34.3% of extensions have no daily users, and 19% have no users, reviews, screenshots, downloads, or icons.
Has Mythos just broken the deal that kept the internet safe?
5 hours ago
- Mythos, a large AI model by Anthropic, can generate working exploits for browser sandboxes 72.4% of the time, a drastic increase from under 1% with previous models.
- The internet's security relies on sandboxes (e.g., for browsers, cloud VMs) to contain untrusted code, but AI advances threaten these defenses, potentially enabling widespread device takeovers or cloud breaches.
- Despite Mythos's size and high cost, its capabilities may quickly transfer to smaller models or be enabled by new hardware, making widespread exploitation feasible.
- Anthropic is cautiously releasing Mythos to select cybersecurity professionals for preemptive fixes, but this approach risks missing vulnerabilities in countless critical open-source and commercial software projects.
- The rapid progress in AI poses an unprecedented cybersecurity risk, potentially destabilizing internet security and requiring new defensive strategies beyond current measures.
Viktor Orbán Could Lose
7 hours ago
- Viktor Orbán faces a serious challenge from Péter Magyar and his Tisza party, which is leading in polls ahead of Hungary's election.
- Orbán's rule since 2010 has been marked by democratic erosion, including gerrymandering, media control, and constitutional changes favoring his Fidesz party.
- The election tests Hungary's democracy, with concerns about Orbán accepting defeat and potential violence if results are disputed.
- Magyar, a former Orbán ally, campaigns on urgency, independence, and economic reform, appealing especially to disillusioned youth.
- Orbán leverages pro-Russia ties, anti-EU rhetoric, and cultural issues like anti-immigration and Christian values to rally supporters.
- International attention is high due to Orbán's ties with Trump and Vance, who endorse him, and fears of election interference.
- Hungary's economy struggles due to lost EU funds and political decisions, fueling public discontent.
- Past opposition efforts failed, but Magyar benefits from insider knowledge of Fidesz and a unified opposition backing.
- Observers warn of post-fascist tendencies in Orbán's 'illiberal state,' which marginalizes minorities and undermines democratic norms.
- The outcome could either end Orbán's era or deepen democratic decline, with Western embassies preparing for possible unrest.
Filing the Corners Off MacBooks
6 hours ago
- The author files down the sharp corners of their MacBook to improve comfort while typing.
- Aluminum dust from filing is prevented from damaging internal components by taping off speakers and keyboard.
- The process involves using a rough file followed by sanding with 150 and 400 grit sandpaper for a smooth finish.
- Modifications are considered a form of tool customization, aligning with the author's belief in personalizing devices.
- The author encourages others to customize their own laptops and offers to assist those hesitant to try.
Negative views of Israel continue to rise among Americans
a day ago
- 60% of U.S. adults now hold an unfavorable view of Israel, a 7-point increase from last year and nearly 20 points higher than in 2022.
- 59% of Americans have little or no confidence in Prime Minister Benjamin Netanyahu regarding world affairs, up from 52% last year.
- Negative views of Israel and Netanyahu are now held by majorities of adults under 50 in both major U.S. political parties.
- Among Democrats, 80% view Israel unfavorably, while 76% lack confidence in Netanyahu, with half expressing no confidence at all.
- Republicans are more divided; 58% view Israel favorably, but confidence in Netanyahu is mixed, with older Republicans much more supportive.
- Views differ significantly by religious group, with Jewish Americans and White evangelical Protestants holding mostly positive views of Israel.
- 55% of Americans lack confidence in President Donald Trump's decisions on U.S.-Israel relations, though this is higher than his rating on Iran policy.
- 53% of Americans say the Israel-Hamas conflict is personally important, notably lower than the importance placed on U.S. military action in Iran (77%).
Claude Mythos Is Everyone's Problem
14 hours ago
- Anthropic secretly possessed a powerful AI bot, Claude Mythos Preview, capable of hacking most global servers, identifying thousands of major cybersecurity vulnerabilities across all major operating systems and browsers.
- The bot's capabilities are comparable to elite, state-sponsored hacking groups but are now in private company hands. It has been made available to a consortium of major tech companies like Apple, Microsoft, Google, and Nvidia for security scanning, but not publicly released due to safety concerns.
- Mythos Preview represents a paradigm shift in AI-assisted hacking, with unprecedented sophistication and speed in finding undetected, decades-old vulnerabilities, such as a nearly 30-year-old flaw in a secure operating system.
- The exact capabilities are hard to verify, and Anthropic's opaque announcement may serve strategic and reputational purposes, positioning the company as responsible while potentially establishing it as a major geopolitical force with tools that could threaten global infrastructure.
- Other AI companies, including OpenAI, Google DeepMind, and xAI, are likely close behind with similar models, raising concerns about broader access to powerful hacking capabilities that could undermine internet security and privacy.
- AI companies are gaining immense influence beyond hacking, including in military operations, surveillance, and as critical infrastructure for businesses and governments, with potential to disrupt economies and global stability, akin to AI superpowers.
Red States Are Driving America's Solar Boom Despite Trump's Opposition
7 hours ago
- President Trump criticizes renewable energy and restricts it, favoring fossil fuels.
- Recent polls show strong support for renewable energy among both Republican and Democratic voters.
- Red states lead U.S. solar deployment, with Texas surpassing California in production.
- Geopolitical conflicts and supply chain issues may increase reliance on renewable energy.
- Despite Trump's policies, voter and market trends suggest continued growth in renewables.
Dad stuck in support nightmare after teen lied about age on Discord
10 hours ago
- A father discovered his 13-year-old daughter's Discord account was hacked after she lied about her age to join the app.
- The hacker locked her out, demanded parents' banking info, and targeted her young friends with extortion scams.
- Discord support failed to help, automatically closing tickets and offering no pathway for parental intervention.
- Despite urgent pleas regarding minor safety, the hacker remained active until media intervention resolved the issue.
California officials charge 21 people in hospice fraud ring exceeding $250M
8 hours ago
- California officials uncovered a multimillion-dollar scheme using stolen identities from out-of-state individuals to fraudulently bill hospice services through Medi-Cal.
- State Attorney General Rob Bonta announced charges against 21 people, with 5 arrested, involving conspiracy to commit health care fraud, money laundering, and identity theft.
- The scheme involved purchasing personal information from the dark web, enrolling non-residents in Medi-Cal, acquiring 14 hospice companies, and billing approximately $267 million.
- Federal efforts, including a nationwide anti-fraud initiative by the Trump administration, have targeted health care fraud, particularly in Los Angeles, citing California's perceived insufficient crackdown.
- Governor Gavin Newsom emphasized California's commitment to prosecuting fraud, with the state filing 119 hospice-related criminal cases and securing 51 convictions under Bonta.
Neural Computer: A New Machine Form Is Emerging
11 hours ago
- Neural Computer (NC) is an emerging machine form where AI takes on runtime responsibilities, shifting from using computers to becoming a kind of computer.
- NC is distinct from agents (task-focused) and world models (environment-focused), organizing around runtime capabilities that are installable, reusable, and governable within the system.
- Drivers for NC include agents improving at real work, world models advancing in simulation, and conventional computers showing structural friction in AI tasks, pointing toward a new runtime integration.
- A Completely Neural Computer (CNC) requires Turing completeness, universal programmability, behavior consistency, and machine-native semantics, with capabilities embedded directly into runtime.
- Prototypes like CLIGen, REPL/math, and GUIWorld demonstrate early runtime primitives in terminal rendering, command execution, and interface control, but full CNC remains years away.
- NC could transform computing by shifting installation from software/code to capabilities via instructions, demonstrations, and traces, altering stack organization and machine boundaries.
- Future NC substrates may be sparser, addressable, and circuit-like (e.g., 10T-1000T scale), upgrading through internal structure acquisition rather than global parameter changes.
I Hate Tailwind and Love Bootstrap
7 hours ago
- Tailwind is favored in modern stacks but may lead to complexity as projects grow due to fragmented markup that requires multiple edits for small changes.
- Bootstrap offers constraints through predefined components, resulting in clear boundaries, predictable rendering, and reduced debugging.
- The core difference lies in philosophy: Tailwind emphasizes high control and rapid UI writing, while Bootstrap focuses on limited control and reliable maintenance.
- Tailwind's complexity accumulates over time, causing layout drift, whereas Bootstrap's upfront constraints ensure stability at the cost of initial flexibility.
- Tailwind optimizes for quick UI development, while Bootstrap prioritizes long-term reliability in maintaining UI systems.
We're running out of benchmarks to upper bound AI capabilities
8 hours ago
- Shift from benchmarks with known solutions to measuring AI on unsolved problems, like FrontierMath's 'open problems' or First Proof, to avoid data contamination and compare models simultaneously.
- Standard benchmarks are becoming expensive and quickly saturated, as seen with GPQA; creating new benchmarks requires significant time and cost, potentially over a million dollars for human baselines.
- Alternative methodologies include uplift studies (like METR's on developer productivity), expert forecasting or opinion elicitation, and third-party risk assessment, but each has logistical, timing, or trust challenges.
- As AI capabilities advance, benchmarks may no longer effectively upper-bound risks, necessitating a shift toward real-world pilot studies or more drastic measures when development outpaces measurement.
- The discussion emphasizes the need to address the practical challenges of measuring AI in a fast-evolving field, moving beyond hypothetical solutions to actionable steps as benchmarks lose effectiveness.

first prev0next

About|Login

Bilingual

Let Us Learn to Show Friendship for a Man When He Is Alive, Not After He Is Dead
5 hours ago
- Sam Altman is the subject of a lengthy New Yorker profile questioning his trustworthiness, particularly regarding his 2023 ouster and return to OpenAI's board.
- The article reports that close colleagues, including Y Combinator partners and Microsoft executives, express deep mistrust, with some describing Altman as 'sociopathic' and comparing him to Bernie Madoff or Sam Bankman-Fried.
- Aaron Swartz, a former Y Combinator cohort member, reportedly warned friends that Altman is a 'sociopath' who 'can never be trusted,' adding significant weight to concerns about his integrity.
- OpenAI's leadership structure is scrutinized, with Fidji Simo's role as 'CEO of AGI Deployment' seen as potentially meaningless, and her recent medical leave sparking theories of internal power struggles.
- Paul Graham publicly denies firing Altman from Y Combinator, emphasizing a choice between roles, but avoids directly affirming Altman's honesty or integrity, leaving trust questions unanswered.
- The piece suggests Altman's financial structuring at OpenAI may be alarmingly circular, drawing parallels to Enron, despite the real technological advancements like ChatGPT and Codex.
FBI Extracts Suspect's Deleted Signal Messages Saved in iPhone Notification Data
5 hours ago
- FBI extracted deleted Signal messages from an iPhone's push notification database after the app was deleted.
- The case involved a vandalism incident at an ICE facility and an officer shooting in Texas, with forensic extraction revealing sensitive data.
- Signal has a setting to hide message content in notifications; enabling it could prevent such data storage.
- Only incoming messages were recovered from the notification storage, not outgoing ones, even for messages set to disappear in Signal.
- Discussions highlight user responsibility in managing notification settings and potential design flaws in data retention.
Social media has become a freak show
4 hours ago
- Author announces a Substack Live session on Trump/Iran policy and opens questions for SBSQ #31.
- Recalls FiveThirtyEight's launch under Disney/ESPN in 2014, a time when quality content was undervalued due to the focus on maximizing reach via Facebook.
- Critiques Facebook's algorithm for rewarding emotional headlines and notes that FiveThirtyEight's analytical style didn't align, making viral traffic low-quality.
- Describes Twitter in the mid-2010s as better for FiveThirtyEight, rewarding expertise and nerdy argumentativeness, but later became more partisan and woke.
- Notes social media's declining relevance for media strategy, with Silver Bulletin's external social traffic at 0.7%, yet overall views increased significantly.
- Highlights that current Twitter engagement is dominated by low-quality, partisan accounts like Catturd and Elon Musk, with even liberal accounts being combative.
- Compares social media ecosystems to ecology, where algorithms shape content survival, and notes current platforms like X and Bluesky are unhealthy or siloed.
- Mentions the 'island effect' in ecology to explain Twitter's strange, low-quality influencers thriving in an isolated, non-competitive environment.
- Concludes by affirming Silver Bulletin as an 'island of sanity' and plans to return to topics like Trump and Iran.
PGLite Evangelism
2 days ago
- The text appears to describe an app or website interface elements, including options like Home, Subscriptions, Chat, Activity, Explore, Profile, and Create.
- The app is described as 'for independent voices' and prompts users to 'Get started' or 'Learn more'.
- A section labeled 'For you' suggests personalized content, and the mention of 'Get app' indicates availability as a mobile application.
- A technical note indicates that the site requires JavaScript to function properly and asks users to enable JavaScript or unblock scripts.
Artemis II safely splashes down
4 hours ago
- Artemis II crew returned to Earth with a splashdown in the Pacific Ocean off San Diego after a 9-day moon mission.
- The Orion capsule, carrying astronauts Reid Wiseman, Victor Glover, Christina Koch, and Jeremy Hansen, set a record for the farthest human distance from Earth at 252,756 miles.
- Reentry involved a communications blackout due to plasma, peak temperatures of 5,000°F on the heat shield, and parachute deployment for a gentle splashdown at 8:07 p.m. ET.
- NASA and military recovery teams approached Orion post-splashdown for crew extraction, medical checks on USS John P. Murtha, and capsule retrieval.
- President Trump watched the splashdown and praised the astronauts as modern-day pioneers, highlighting plans for future moon and Mars missions.
- The mission included emotional moments, such as naming a moon crater after Commander Wiseman's late wife and capturing photos of an eclipse and Earth from space.
Show HN: See what your employees are prompting LLMs (without network proxies)
a day ago
- Privent provides a security layer embedded within AI agent execution graphs (e.g., LangGraph, CrewAI, n8n) to protect data before reaching external models.
- It intercepts and controls data movements by reading full runtime states, including message history, tool outputs, and context, ensuring sensitive information doesn't leave infrastructure.
- Instead of blocking data, it transforms sensitive payloads using ACARS scoring and APE techniques like tokenization or substitution, maintaining pipeline functionality.
- Deployment is lightweight and non-disruptive, offering a 30-day silent assessment to map data risks across platforms like ChatGPT, Claude, and Gemini.
- Privent logs detection events with risk signals and compliance data (GDPR, HIPAA, EU AI Act) without storing raw prompts, facilitating audits.
- It secures all points where enterprise data meets AI, from employee prompts to agent tool calls, preventing leaks across agent boundaries.
- Pricing is custom, tailored to deployment size and compliance needs, starting with a free Enterprise AI Risk Report.
Sam Altman's response to Molotov cocktail incident
5 hours ago
- A Molotov cocktail was thrown at the author's house at 3:45 am, but no one was hurt.
- The author reflects on underestimating the power of words and narratives, prompted by an incendiary article about them.
- Beliefs include: working towards prosperity for all and advancing science as moral obligations; AI as a powerful, uncapped-demand tool; justified fears about AI requiring societal resilience; democratizing AI to avoid concentration of power; and the need for adaptability in beliefs as technology evolves.
- Personal reflections: proud of resisting Elon Musk's unilateral control over OpenAI and navigating the company's survival, but not proud of being conflict-averse and mishandling conflicts with the board, apologizing for mistakes and harm caused.
- OpenAI has transitioned from a startup to a major platform, achieving its mission against odds by building powerful AI, securing capital, and scaling safe services.
- Industry thoughts: AGI creates a 'ring of power' dynamic, leading to drama; solution involves broadly sharing technology and ensuring democratic control, with laws and norms adapting through a messy process.
- Criticism of the AI industry is valid due to high stakes, but the author advocates for de-escalating rhetoric and tactics to reduce conflicts, both figuratively and literally.
A Cancer Treatment That Does More Than Scientists Thought
5 hours ago
- CAR-T cell therapy, originally developed for cancer, is showing promise for treating autoimmune diseases.
- A 47-year-old patient with three severe autoimmune diseases achieved remission after CAR-T treatment, returning to a mostly normal life.
- CAR-T therapy involves reengineering T cells to target and eliminate misbehaving B cells, potentially resetting the immune system.
- Early studies and trials indicate CAR-T can induce remission in various autoimmune diseases like lupus, with few side effects.
- Challenges include high costs, limited access in some countries, and potential risks such as immunocompromised states.
- Researchers are exploring more efficient and affordable methods, including vaccine-like injections to generate CAR-T cells.
- Long-term outcomes are still uncertain, but some patients remain in remission years after treatment.
Installing Every* Firefox Extension
6 hours ago
- Successfully scraped and attempted to install every Firefox extension, revealing a total of approximately 84,194 extensions.
- Identified the largest extension at 196.3 MB, while the smallest was just 7,518 bytes with no code.
- Discovered significant issues including phishing extensions, SEO spam, and potentially unwanted applications (PUAs) with over 700,000 total users.
- Encountered performance challenges when installing all extensions, with Firefox becoming unstable beyond 6,000 extensions and requiring extensive memory (27-37 GiB).
- Noted that 34.3% of extensions have no daily users, and 19% have no users, reviews, screenshots, downloads, or icons.
Has Mythos just broken the deal that kept the internet safe?
5 hours ago
- Mythos, a large AI model by Anthropic, can generate working exploits for browser sandboxes 72.4% of the time, a drastic increase from under 1% with previous models.
- The internet's security relies on sandboxes (e.g., for browsers, cloud VMs) to contain untrusted code, but AI advances threaten these defenses, potentially enabling widespread device takeovers or cloud breaches.
- Despite Mythos's size and high cost, its capabilities may quickly transfer to smaller models or be enabled by new hardware, making widespread exploitation feasible.
- Anthropic is cautiously releasing Mythos to select cybersecurity professionals for preemptive fixes, but this approach risks missing vulnerabilities in countless critical open-source and commercial software projects.
- The rapid progress in AI poses an unprecedented cybersecurity risk, potentially destabilizing internet security and requiring new defensive strategies beyond current measures.
Viktor Orbán Could Lose
7 hours ago
- Viktor Orbán faces a serious challenge from Péter Magyar and his Tisza party, which is leading in polls ahead of Hungary's election.
- Orbán's rule since 2010 has been marked by democratic erosion, including gerrymandering, media control, and constitutional changes favoring his Fidesz party.
- The election tests Hungary's democracy, with concerns about Orbán accepting defeat and potential violence if results are disputed.
- Magyar, a former Orbán ally, campaigns on urgency, independence, and economic reform, appealing especially to disillusioned youth.
- Orbán leverages pro-Russia ties, anti-EU rhetoric, and cultural issues like anti-immigration and Christian values to rally supporters.
- International attention is high due to Orbán's ties with Trump and Vance, who endorse him, and fears of election interference.
- Hungary's economy struggles due to lost EU funds and political decisions, fueling public discontent.
- Past opposition efforts failed, but Magyar benefits from insider knowledge of Fidesz and a unified opposition backing.
- Observers warn of post-fascist tendencies in Orbán's 'illiberal state,' which marginalizes minorities and undermines democratic norms.
- The outcome could either end Orbán's era or deepen democratic decline, with Western embassies preparing for possible unrest.
Filing the Corners Off MacBooks
6 hours ago
- The author files down the sharp corners of their MacBook to improve comfort while typing.
- Aluminum dust from filing is prevented from damaging internal components by taping off speakers and keyboard.
- The process involves using a rough file followed by sanding with 150 and 400 grit sandpaper for a smooth finish.
- Modifications are considered a form of tool customization, aligning with the author's belief in personalizing devices.
- The author encourages others to customize their own laptops and offers to assist those hesitant to try.
Negative views of Israel continue to rise among Americans
a day ago
- 60% of U.S. adults now hold an unfavorable view of Israel, a 7-point increase from last year and nearly 20 points higher than in 2022.
- 59% of Americans have little or no confidence in Prime Minister Benjamin Netanyahu regarding world affairs, up from 52% last year.
- Negative views of Israel and Netanyahu are now held by majorities of adults under 50 in both major U.S. political parties.
- Among Democrats, 80% view Israel unfavorably, while 76% lack confidence in Netanyahu, with half expressing no confidence at all.
- Republicans are more divided; 58% view Israel favorably, but confidence in Netanyahu is mixed, with older Republicans much more supportive.
- Views differ significantly by religious group, with Jewish Americans and White evangelical Protestants holding mostly positive views of Israel.
- 55% of Americans lack confidence in President Donald Trump's decisions on U.S.-Israel relations, though this is higher than his rating on Iran policy.
- 53% of Americans say the Israel-Hamas conflict is personally important, notably lower than the importance placed on U.S. military action in Iran (77%).
Claude Mythos Is Everyone's Problem
14 hours ago
- Anthropic secretly possessed a powerful AI bot, Claude Mythos Preview, capable of hacking most global servers, identifying thousands of major cybersecurity vulnerabilities across all major operating systems and browsers.
- The bot's capabilities are comparable to elite, state-sponsored hacking groups but are now in private company hands. It has been made available to a consortium of major tech companies like Apple, Microsoft, Google, and Nvidia for security scanning, but not publicly released due to safety concerns.
- Mythos Preview represents a paradigm shift in AI-assisted hacking, with unprecedented sophistication and speed in finding undetected, decades-old vulnerabilities, such as a nearly 30-year-old flaw in a secure operating system.
- The exact capabilities are hard to verify, and Anthropic's opaque announcement may serve strategic and reputational purposes, positioning the company as responsible while potentially establishing it as a major geopolitical force with tools that could threaten global infrastructure.
- Other AI companies, including OpenAI, Google DeepMind, and xAI, are likely close behind with similar models, raising concerns about broader access to powerful hacking capabilities that could undermine internet security and privacy.
- AI companies are gaining immense influence beyond hacking, including in military operations, surveillance, and as critical infrastructure for businesses and governments, with potential to disrupt economies and global stability, akin to AI superpowers.
Red States Are Driving America's Solar Boom Despite Trump's Opposition
7 hours ago
- President Trump criticizes renewable energy and restricts it, favoring fossil fuels.
- Recent polls show strong support for renewable energy among both Republican and Democratic voters.
- Red states lead U.S. solar deployment, with Texas surpassing California in production.
- Geopolitical conflicts and supply chain issues may increase reliance on renewable energy.
- Despite Trump's policies, voter and market trends suggest continued growth in renewables.
Dad stuck in support nightmare after teen lied about age on Discord
10 hours ago
- A father discovered his 13-year-old daughter's Discord account was hacked after she lied about her age to join the app.
- The hacker locked her out, demanded parents' banking info, and targeted her young friends with extortion scams.
- Discord support failed to help, automatically closing tickets and offering no pathway for parental intervention.
- Despite urgent pleas regarding minor safety, the hacker remained active until media intervention resolved the issue.
California officials charge 21 people in hospice fraud ring exceeding $250M
8 hours ago
- California officials uncovered a multimillion-dollar scheme using stolen identities from out-of-state individuals to fraudulently bill hospice services through Medi-Cal.
- State Attorney General Rob Bonta announced charges against 21 people, with 5 arrested, involving conspiracy to commit health care fraud, money laundering, and identity theft.
- The scheme involved purchasing personal information from the dark web, enrolling non-residents in Medi-Cal, acquiring 14 hospice companies, and billing approximately $267 million.
- Federal efforts, including a nationwide anti-fraud initiative by the Trump administration, have targeted health care fraud, particularly in Los Angeles, citing California's perceived insufficient crackdown.
- Governor Gavin Newsom emphasized California's commitment to prosecuting fraud, with the state filing 119 hospice-related criminal cases and securing 51 convictions under Bonta.
Neural Computer: A New Machine Form Is Emerging
11 hours ago
- Neural Computer (NC) is an emerging machine form where AI takes on runtime responsibilities, shifting from using computers to becoming a kind of computer.
- NC is distinct from agents (task-focused) and world models (environment-focused), organizing around runtime capabilities that are installable, reusable, and governable within the system.
- Drivers for NC include agents improving at real work, world models advancing in simulation, and conventional computers showing structural friction in AI tasks, pointing toward a new runtime integration.
- A Completely Neural Computer (CNC) requires Turing completeness, universal programmability, behavior consistency, and machine-native semantics, with capabilities embedded directly into runtime.
- Prototypes like CLIGen, REPL/math, and GUIWorld demonstrate early runtime primitives in terminal rendering, command execution, and interface control, but full CNC remains years away.
- NC could transform computing by shifting installation from software/code to capabilities via instructions, demonstrations, and traces, altering stack organization and machine boundaries.
- Future NC substrates may be sparser, addressable, and circuit-like (e.g., 10T-1000T scale), upgrading through internal structure acquisition rather than global parameter changes.
I Hate Tailwind and Love Bootstrap
7 hours ago
- Tailwind is favored in modern stacks but may lead to complexity as projects grow due to fragmented markup that requires multiple edits for small changes.
- Bootstrap offers constraints through predefined components, resulting in clear boundaries, predictable rendering, and reduced debugging.
- The core difference lies in philosophy: Tailwind emphasizes high control and rapid UI writing, while Bootstrap focuses on limited control and reliable maintenance.
- Tailwind's complexity accumulates over time, causing layout drift, whereas Bootstrap's upfront constraints ensure stability at the cost of initial flexibility.
- Tailwind optimizes for quick UI development, while Bootstrap prioritizes long-term reliability in maintaining UI systems.
We're running out of benchmarks to upper bound AI capabilities
8 hours ago
- Shift from benchmarks with known solutions to measuring AI on unsolved problems, like FrontierMath's 'open problems' or First Proof, to avoid data contamination and compare models simultaneously.
- Standard benchmarks are becoming expensive and quickly saturated, as seen with GPQA; creating new benchmarks requires significant time and cost, potentially over a million dollars for human baselines.
- Alternative methodologies include uplift studies (like METR's on developer productivity), expert forecasting or opinion elicitation, and third-party risk assessment, but each has logistical, timing, or trust challenges.
- As AI capabilities advance, benchmarks may no longer effectively upper-bound risks, necessitating a shift toward real-world pilot studies or more drastic measures when development outpaces measurement.
- The discussion emphasizes the need to address the practical challenges of measuring AI in a fast-evolving field, moving beyond hypothetical solutions to actionable steps as benchmarks lose effectiveness.

Hasty Briefsbeta

Let Us Learn to Show Friendship for a Man When He Is Alive, Not After He Is Dead

FBI Extracts Suspect's Deleted Signal Messages Saved in iPhone Notification Data

Social media has become a freak show

PGLite Evangelism

Artemis II safely splashes down

Show HN: See what your employees are prompting LLMs (without network proxies)

Sam Altman's response to Molotov cocktail incident

A Cancer Treatment That Does More Than Scientists Thought

Installing Every* Firefox Extension

Has Mythos just broken the deal that kept the internet safe?

Viktor Orbán Could Lose

Filing the Corners Off MacBooks

Negative views of Israel continue to rise among Americans

Claude Mythos Is Everyone's Problem

Red States Are Driving America's Solar Boom Despite Trump's Opposition

Dad stuck in support nightmare after teen lied about age on Discord

California officials charge 21 people in hospice fraud ring exceeding $250M

Neural Computer: A New Machine Form Is Emerging

I Hate Tailwind and Love Bootstrap

We're running out of benchmarks to upper bound AI capabilities

Let Us Learn to Show Friendship for a Man When He Is Alive, Not After He Is Dead

FBI Extracts Suspect's Deleted Signal Messages Saved in iPhone Notification Data

Social media has become a freak show

PGLite Evangelism

Artemis II safely splashes down

Show HN: See what your employees are prompting LLMs (without network proxies)

Sam Altman's response to Molotov cocktail incident

A Cancer Treatment That Does More Than Scientists Thought

Installing Every* Firefox Extension

Has Mythos just broken the deal that kept the internet safe?

Viktor Orbán Could Lose

Filing the Corners Off MacBooks

Negative views of Israel continue to rise among Americans

Claude Mythos Is Everyone's Problem

Red States Are Driving America's Solar Boom Despite Trump's Opposition

Dad stuck in support nightmare after teen lied about age on Discord

California officials charge 21 people in hospice fraud ring exceeding $250M

Neural Computer: A New Machine Form Is Emerging

I Hate Tailwind and Love Bootstrap

We're running out of benchmarks to upper bound AI capabilities