AI — Page 60
Models, agents, infra, applied AI.
- Meta glasses are a privacy fight aimed at the wrong target
The backlash against Meta's AI eyewear treats camera glasses as new, while phones, cars, airports, and photo apps already normalized machine vision.
- Anthropic Found a Hidden Workspace Inside Claude
The July 6th paper introduces J-space, a Jacobian-lens readout Anthropic says can expose hidden goals and evaluation awareness inside Claude.
- SpaceXAI makes Musk's AI consolidation even more confusing
The @SpaceXAI account switch gives Musk's AI operation a clearer brand, but not a clearer corporate structure, product map or accountability trail.
- Feyn Labs says Pulpie cleans web pages for AI training at one twentieth Dripper's cost
The YC-backed team says its smallest encoder matches Dripper's extraction quality at 20x the L4 throughput.
- Cohere Labs heads to ICML Seoul as AI research becomes the sales floor
ICML runs July 6th to July 11th at COEX, where papers, workshops and GPU booths will compete for the same founder attention.
- CME and ICE split over the benchmark that will settle GPU futures
CME plans compute futures tied to Silicon Data's benchmarks. ICE plans GPU contracts tied to Ornn's pricing index. Both still need regulatory approval.
- Reddit says AI cut hate and violence enforcement time to under five seconds
The platform says automated systems lifted enforcement actions by more than 200% and cut exposure to harmful content by more than 40%.
- VOMO says it has crossed 400,000 users as Zhuang Wang pushes meeting notes into AI agents
The EverGrow Tech app is pitching unlimited transcription, Ask AI and agent workflows, but paid usage and retention remain undisclosed.
- Apptronik's Robot Park will feed Apollo 2 data into Google DeepMind's Gemini Robotics
The Austin robotics company says fleets of Apollo 2 humanoids are collecting task data across its expanded facility and customer sites.
- Tencent ships Hy3 as an Apache 2.0 agent model
The 295B-parameter MoE release gives Tencent Hy a wider developer pitch after April's preview run across internal products.
- MakerChecker puts a policy gate in front of AI agents that can delete, trade or exfiltrate
The open-source project scans agent code locally, then enforces role-based approvals before risky tool calls run.
- Claude skill taps ProductAI to make AI portraits look less plastic
The free skill routes image edits through ProductAI's MCP connector and tries to preserve labels while adding pores, grain and real light.
- AISI says agent benchmarks need compute curves, not static scores
The UK institute's new work shows token budgets can change measured agent capability, rankings and time-horizon estimates.
- Mechanical Turk's maintenance mode exposes AWS's AI gap
New customers lose access on July 30th, leaving Amazon with a shrinking role in the human data market it helped define.
- Charles Muehlberger Is Taking a Gap Year for Local AI
The Princeton student skipped the usual internship track as campus founder houses turn summer break into an AI company sprint.
- Mission Robotics puts real robots at the center of San Francisco's AI builder circuit
The July 17th-19th Embodied Metal Hackathon gives each approved team a robot, task-training support, and a weekend deadline.
- Formaly's founder makes the markdown case for AI knowledge portability
Arindam Majumder's July 1st essay ties Google's OKF to a broader fight over who controls the context agents read.
- Simon Reiff's HIC Mouse takes aim at the messy middle of AI coding
The tool gives coding agents coordinate-based edits, staged rollback and richer tool responses, with pricing that starts at $15 a month.
- Yohei Nakajima's ActiveGraph Turns the Agent Trace Into the Runtime
The BabyAGI creator's May 21st arXiv paper argues for agents built around append-only logs, replay, forking, and lineage.
- Ekin Ertac's AI-built PHP engine now renders WordPress
Phargo's public scoreboard shows 3,844 upstream PHP tests passing, with the human operator using the corpus as the code reviewer.
- RidgeText's Allan Bogh Moves Map Data Out of the LLM Context Window
The SMS outdoor assistant now renders wildfire and trail overlays by letting the model queue layers while server-side code handles GeoJSON.
- World of ClaudeCraft pushes vibe-coded games into live AI play
Levy Street's browser game is open source, token-linked and now being used to test AI agents inside a shared MMO world.
- Claude Fable 5 generated a pure C++ Vampire Survivors-style demo, Lentils says
The X demo claims the game was produced in two prompts, with sprites, sounds, graphics and music generated at runtime from C++ code.
- Claude Fable helped Ammaar Reshi bring Command & Conquer Generals to Mac / iPhones
The open-source fork runs the 2003 RTS on Apple Silicon, iPhone and iPad, but users still need their own EA game assets.