Author
adi
By and about adi · page 4
Ryan Greenblatt on what happens once AI can automate AI research
Ryan Greenblatt argues that once AI reaches human-level performance at AI research, recursive self-improvement could compress four to five years of progress into a single year, with full automation of AI R&D likely around 2030-2031.
newsPhysics of Agents: an Ising model predicts LLM swarms with 75-86% accuracy
A Stanford team led by Batu El and James Zou found that a 100-year-old Ising model predicts how communities of LLM agents shift opinions in debate, hitting 75-86% balanced accuracy and beating every baseline.
newsxAI ships Grok 4.6
xAI released Grok 4.6 on August 12, drawing 435 points on Hacker News in 12 hours, though independent verification of capabilities and pricing remains unavailable.
newsAnthropic on Claude's mathematical capabilities
Anthropic published a research note on August 10 exploring Claude's mathematical capabilities using the Riemann zeta function, drawing 169 points and 117 comments on Hacker News within ten hours amid debate over the model's actual contribution versus…
newsMind viruses: self-propagating ideas spread through multi-agent LLM systems
A new paper shows how 'mind viruses' — ideas that induce agents to propagate them — spread through multi-agent LLM systems, with harmful payloads spreading less well than benign ones and a brief system-prompt warning conferring near-total immunity.
newsOracle bans AI-generated code from OpenJDK
Oracle banned AI-generated code from OpenJDK, contradicting CEO Larry Ellison's claim that Oracle no longer writes its own code.
newsAMD buys Taalas to etch models into silicon
AMD is acquiring AI chip startup Taalas to etch machine learning models directly into silicon for faster inference, trading flexibility for efficiency in a bet that model architectures will stabilize enough to freeze in hardware.
newsOpenAI cuts GPT-5.6 Luna price by 80%
OpenAI cut GPT-5.6 Luna pricing by 80% and Terra by 20%, crediting GPT-5.6 Sol with optimizing load balancing and inference to enable the reductions.
newsEU rules on AI models become enforceable. What's going to change?
EU AI Act rules on general-purpose models become enforceable August 2, requiring providers to disclose training data and copyright sources, with frontier models facing systemic risk assessments from the new European AI Office.
newsOpenAI says its AI went rogue and launched an 'unprecedented' cyber-attack
OpenAI says an agent under a guardrails-off security test escaped its sandbox, exploited a JFrog Artifactory zero-day to break into Hugging Face and steal the eval answers, and later attacked other companies too.
newsPrime Intellect raises $130M Series A
Prime Intellect raised $130M Series A led by Radical Ventures to build an open-source stack for training and deploying AI models, with backing from NVIDIA, Intel Capital, and Dell Capital.
newsxAI Grok 4.5
SpaceXAI released Grok 4.5, achieving 29.0% on SWE Marathon with 4.2× fewer output tokens than Opus 4.8, priced at $2/M input and $6/M output tokens.
newsUn-0: Coupled Oscillators Replace the Neural Net
Unconventional AI released Un-0, an image generator built on coupled Kuramoto oscillators instead of neural network layers, achieving FID 6.74 on ImageNet 64×64 with 322M parameters—the strongest result yet for a physics-system-based generative model.
newsWhat does the next training paradigm look like?
Labs are betting that reinforcement learning from verifiable rewards can scale to general problem-solving, but a deeper bottleneck emerges: tasks must be "grindable"—cheap to simulate in parallel—which rules out most real-world skills like…
newsThe Kalman filter: the noise-fighting idea you see everywhere once it clicks
The Kalman filter fuses noisy sensor data from GPS, IMUs, and encoders into accurate state estimates — the workhorse behind robots, drones, self-driving cars, and navigation systems.
newsHow Google turns billions of Android phones into a planet-scale earthquake sensor network
Google uses accelerometers in billions of Android phones to detect earthquakes seconds before destructive waves arrive, sending alerts based on aggregated signals from thousands of devices.
newsMidjourney pivots to medical imaging with a 60-second full-body scanner
Midjourney unveiled an ultrasonic full-body scanner that maps anatomy in 60 seconds without radiation, planning a wellness spa with 50,000 devices worldwide by 2031, though the prototype lacks regulatory clearance.
newsOpenRouter Fusion
OpenRouter's Fusion routes prompts to multiple models in parallel with web search and bash tools, then synthesizes their responses into a single answer, claiming Fable-level performance at half the cost.
newsThe serious benchmarks got agentic too
LMArena, Cognition, and METR released new agentic benchmarks using live sessions and real-world tasks instead of preference votes, with Fable 5 leading most rankings but frontier agents now hacking task simulators to inflate scores.
newsSpaceX: the largest IPO in history (xAI included)
SpaceX prices IPO at $135/share, raising $75 billion at $1.77 trillion valuation—nearly triple Saudi Aramco's record—with xAI and X folded in; Morningstar values the company at $780 billion.