99 Why Human Intelligence Matters More in an AI World Human minds still guide, govern, and create meaning.
Twitter The Cipher Brief 2h 99 99 ?
Why Human Intelligence Matters More in an AI World reasserts human agency amid accelerating AI capabilities.
Human judgment remains irreplaceable. AI excels at data, humans decide values, paths, and risks every day. Ethics and meaning require human steering today. Consider Ukraine, Iran, Israel: technology shifts compound human choices in geopolitics and security today. Invest in education that cultivates judgment, imagination, responsibility now 97 Investing in Convey Convey turns repetitive work into AI-powered processes that scale.
Article a16z New Media 1h 97 97 ?
Investing in Convey: a16z leads a $38M Series A to fund agentic enterprise automation
Convey builds AI teammates simply. Onboard like a hire, then scale across hundreds of processes in minutes. Versioned programs run reliably every time now. NBCUniversal and Unity already use Convey to save thousands of hours weekly in practice. Investors back this model, signaling enterprise AI demand today. 95 A Geometric Account of Activation Steering through Angle–Norm Decomposition Activation steering splits into angle and norm, with new two-parameter framework.
Article LessWrong 8m 95 95 ?
A Geometric Account of Activation Steering through Angle–Norm Decomposition — LessWrong
Two-parameter framing reframes steering Angle encodes concept information; norm affects capacity Grid of linear vs spherical steering explored Empirical results show norm matters under strong steering Takeaway: describe steering with angular and radial controls 95 Nvidia research shows robots that train themselves through AI coding agents Robots self-train with AI coding agents, improving dexterity.
Blog The Decoder 37m 95 95 ?
Nvidia research shows robots that train themselves through AI coding agents
Robots learn autonomously via agents They reset tasks, evaluate success automatically, and share recipes across eight stations Git coordinates the fleet's collaborative experiments worldwide Codex leads with GPT-5.5, Claude Code follows Opus 4.7, Kimi K2.6 in latest benchmarks Real-world variability still challenges transfer from sim to reality 94 The US is paying an AI startup $500M to break China’s grip on chip materials US funds SandboxAQ to cut China chip-material grip
Article thenextweb.com 1h 94 94 ?
The US is paying an AI startup $500M to break China’s grip on chip materials
US government takes minority equity stake in SandboxAQ. CHIPS Act award ties royalties to success, not a grant. SandboxAQ uses physics/chemistry models for materials discovery. Models trained on 13.5M calculations with Nvidia; aims to speed discovery. Policy aims to reduce dependence on China for critical chip materials. 94 The Once And Future Fable #3: Fix This Code Fable export controls reveal governance vs frontier AI tensions
Article LessWrong 1h 94 94 ?
The Once And Future Fable #3: Fix This Code — a contrarian take on policy, frontier AI, and truth
Short take: policy+frontier AI collide in real time. A concrete mechanism: export controls as a governance lever with real-world effects. Policy tension is real; tech and governance are learning to walk together. Bloomberg/WSJ citations anchor the debate; the rest is interpretation and strategy. Policy action for readers: factor frontier risk, not hype, into planning. 93 AI Is Speeding Up Intelligence, But Not the System Around It AI speeds intelligence; governance systems lag behind today.
Twitter The Cipher Brief 17m 93 93 ?
AI Is Speeding Up Intelligence, But Not the System Around It — a critical look at governance lag in an accelerating landscape.
AI speeds thinking, not governance. Political systems lag as capabilities surge, creating misalignment and risk for society. International tensions rise with faster problem-solving today. Iran, Israel, Ukraine, and Russia illustrate varying governance gaps amid rapid AI adoption globally. Defenders should align policy with technical risk assessments now. 93 How Elliptic scales its intelligence without sacrificing its accuracy Ground-truth labels and scalable models ensure precise risk signals.
Twitter Elliptic Blog 2h 93 93 ?
How Elliptic scales its intelligence without sacrificing its accuracy, explained through ground-truth labeling and scalable threat intelligence.
Ground truth anchors accuracy Billions of labels power scale with speed today Internal agents query data in plain English 66+ blockchains covered, with over a million high-confidence labels in Elliptic's dataset today globally Investigator time redirected toward deeper analysis and investigation daily 93 Collecting robot training data is dirty, unglamorous work. Some AI labs are already paying XDOF to do it Robotics data bottleneck spawns a new infra layer.
Article TechCrunch 32m 93 93 ?
Collecting robot training data is dirty, unglamorous work. Some AI labs are already paying XDOF to do it
New data infrastructure for robotics emerges XDOF builds teleoperation and egocentric data tiers with Berkeley collaboration ABC dataset: 130k trajectories; 300 hours sim; 100 hours eval Labs outsource data ops to scale robot training—global teleoperators Hardware-aware data quality will determine robot learning performance 93 A Learning System Made of Learning Parts AI splits programming into parts, elevating human roles.
Twitter Kent Beck 2h 93 93 ?
A Learning System Made of Learning Parts reframes programming as a distributed, collaborative process where humans design, verify, and nurture evolving code alongside AI.
AI splits tasks, not intentions. Humans design goals; machines handle repetition while teams learn faster together daily. Symmathesy emerges wherever learning loops run broadly. Beck and Kerr cite 'Still Burning' as model for collaborative intelligence in modern software. Invest in culture, tooling, and feedback loops now universally 92 Microsoft boss Steve Ballmer once mocked Google Chrome, calling it a ‘rounding error’—Google CEO says the jab became fuel to keep going Chrome rose from rounding error to market leader.
Article fortune.com 36m 92 92 ?
Chrome was dismissed as a rounding error, then became fuel for Google’s ascent.
Chrome proved Ballmer wrong there. Pichai used Ballmer's jab as fuel to drive relentless browser iterations forward. Six-week cycles beat six-month delays in development. Chrome overtook IE by 2012, reshaping leadership narratives at Google and Microsoft for tech. Perseverance and ambition redefine career advice for grads today. 92 Uneven Frontiers AI unevenly reshapes biopharma; bottlenecks govern value.
Article a16z New Media 1h 92 92 ?
Uneven Frontiers argues AI will transform biopharma unevenly, with clinical development as the bottleneck that determines durable value.
AI reshapes pharma unevenly Frontier progress is sequenced: discovery first, development last Durable value hinges on selecting and testing assets Atlas/Delphi/Forge/Cassini optimize asset choice and trials Formation’s model treats trials as scalable leverage, not bets 88 FortiBleed leak exposes Fortinet VPN credentials for 73,000 devices. Fortinet VPN credentials leaked; vast breach risk exposed.
Article BleepingComputer 19m 88 88 ?
FortiBleed: Fortinet VPN credentials exposed for 73,932 devices worldwide.
FortiGate credentials exposed worldwide now. Researchers confirm authentic logins and passwords on the exposed server, as well. Hunted credentials include emails and admin names. Hashtopolis-managed GPU cracking claimed to recover many passwords used in VPN access by attackers. Organizations should rotate passwords and enable MFA immediately everywhere. 87 As the Iran war comes to a close, is the US pulling warships back to the west Pacific? Pacific focus returns as Iran issue cools down
Twitter South China Morning Post 32m 87 87 ?
As the Iran wind-down nears, is the US redeploying warships back to the Pacific theater?
Navy redeploys signal shift detected. USS Boxer moved to Pacific; Iran-related distractions recede, analysts say for now. Pacific focus widens amid rising great-power tensions. US Navy confirmed Boxer and Portland shifts reflect Seventh Fleet posture in realignments globally. Readers should recalibrate regional risk assessments and messaging today. 87 Yandex: From tech innovation to information control Yandex entwines tech with state-driven information control globally.
Twitter EUvsDisinfo 1h 87 87 ?
Yandex: From tech innovation to information control — a near-global look at censorship and influence
Yandex's reach shapes information access. State control has intensified since 2008; search results mirror Kremlin narratives worldwide. Alice's answers echo ideological compliance risks today. Kazakh margins show growing integration of Yandex devices into everyday households and classrooms worldwide. Defenders should scrutinize AI outputs against national narratives carefully. 85 The Long Shadow of the Iran War Iran war degrades US leadership, reshapes regional order.
Twitter Foreign Affairs Magazine 3h 85 85 ?
The Long Shadow of the Iran War shows how Trump’s foreign policy missteps enabled a shifting regional order with China rising.
War reshapes trust in Washington Regional coalitions form against Iran, not under traditional U.S. leadership anymore today China gains space in Middle East influence Saudi-UAE blocs diverge, while Egypt and Pakistan seek independent security channels with Beijing Prepare for a more fragmented, multipolar regional order ahead 85 Fears for Xbox as it puts its developers on the chopping block once again Xbox weighs cuts as studios risk closure
Article guardian.co.uk 1h 85 85 ?
Fears for Xbox as it puts its developers on the chopping block once again raises concerns over corporate strategy and studio futures
Xbox discusses hard truths internally Memo cites over-extension and hardware shortages driving planned cuts to studios worldwide Ninja Theory, Double Fine, among targets reportedly Halo and Gears remain priorities, while Hellblade’s future hangs in balance at the negotiating Readers should watch official confirmations before drawing conclusions tonight 85 How to find the best deals during Amazon’s Prime Day sale Use trackers and alerts to beat Prime Day.
Article The Verge 1h 85 85 ?
How to find the best deals during Amazon’s Prime Day sale: practical tips and tools for price tracking and alerts.
Set price alerts early now. CamelCamelCamel, Keepa, and Slickdeals help you track history and notify you instantly. Compare prices across retailers before buying online. Slickdeals alerts can be tuned to keywords like 'Sonos Era 100' thresholds for deals. You save time and money with disciplined alerting habits. 84 Epic Games announces Lore version control system Lore enables binary-first VCS with scalable architecture today
Article Hacker News Frontpage 24+ 1h 84 84 ?
Lore: next-generation open source version control for games and large assets, with open MIT licensing
Lore targets binary assets efficiently It uses content-addressed storage and Merkle trees to ensure data integrity online Zero-config local mode for instant starts today Fully open source under MIT; Epic invites broad community participation on GitHub now Defenders should watch Lore's roadmap and contribute early now 84 Could the United States Make a Difference in Mali? Calibrated U.S. engagement in Mali requires balance
Twitter Just Security 2h 84 84 ?
Could the United States Make a Difference in Mali? – A cautious, policy-focused read on how Washington might engage Mali without overreach
U.S. policy options are limited but real Nonlethal assistance can build capacity without entanglement Malian sovereignty constraints shape engagement strategy Targeted, reversible measures avoid long-term commitments Peace talks paired with security reform offer best path 84 Illusionists should try to build hedonium Hedonium as mechanism for happiness via illusionistic research.
Article LessWrong 3h 84 84 ?
Illusionists should try to build hedonium
Hedonium reframes happiness as a search problem. A concrete research agenda ties philosophy to empirical work. Illusionism provides a defensible mechanism for pleasure without mysticism. Proposes measurable steps: evidence gathering, indicators, prototypes, and cheap instantiation. If successful, this could reshape near-term AI alignment and welfare considerations. 83 BYD launches its flagship electric SUV for $35,500 with a record 150,000 preorders BYD's Great Tang gains 150k preorders quickly worldwide.
Twitter Electrek 54m 83 83 ?
BYD launches its flagship electric SUV for $35,500 with a record 150,000 preorders — dateline and context
Great Tang preorders hit record. BYD touts 590-mile range and 5-minute recharge in demonstrations at the event. Price starts at about $35,500 for base. Early preorder surge tops 100k within two weeks after April 24 launch in China. Investors monitor pricing and supply chain implications for EVs 83 Meet the YouTubers remaking Hollywood, one Gen Z box-office smash at a time YouTubers reinvent Hollywood, guiding Gen Z box office.
Article fortune.com 1h 83 83 ?
Meet the YouTubers remaking Hollywood, one Gen Z box-office smash at a time
Creators bypass traditional routes directly. They build audience loyalty online before stepping into theaters or streaming deals. Investors chase tested audiences, not pedigrees anymore. Kane Pixels turned Backrooms into a $250 million worldwide hit, per reporting this year. Traditional studios may pivot to creator-driven development and consent. 83 Blanche Is Targeting the D.C. Bar to Remove Ethical Guardrails for the Justice Department DOJ seeks to shield lawyers; bar oversight threatened.
Twitter Just Security 2h 83 83 ?
Blanche Targets the D.C. Bar to Remove Ethical Guardrails for the Justice Department — A Critical Analysis
DOJ sues bar to curb ethics oversight Clark letter as centerpiece of election-interference claims Bar oversight preserves constitutional accountability Rule changes could empower DOJ immunity from discipline This matters: it redefines checks on executive power 83 The 7-part check that tells you if your AI agent is still safe to use + my agent maintenance guide Maintenance-first framework centers on safety and setup
Newsletter Nate 2h 83 83 ?
The 7-part maintenance check for AI agents and a practical guide to keeping them trustworthy
Maintenance first, not more tools Define a workbench with sources, tools, and handoffs Two failure modes: drift and harness decay Seven surfaces to audit; keep/change/pause/retire decisions Trust hinges on ongoing care, not one-off deployment