Seung Jung Reporter
Seung Jung · Articles (10)
299 articles
AI & Machine Learning
OpenAI Asks California to Regulate It Harder Than SB 53 Currently Does
A short LinkedIn post reverses the company's position on the frontier AI law and endorses states as the drafting floor for a future national standard.
AI & Machine Learning
Inherent's Faraday Agent Replicates Research Papers on a 27B Model
The London lab, founded by Google DeepMind alumni, says its agent beat far larger frontier systems at reproducing published findings.
AI & Machine Learning
Prime Intellect Ran 153 Autonomous Jobs on 18 Models. The Cheapest Winner Used 3% of the Tokens
A public leaderboard for the nanoGPT optimizer speedrun turns agent efficiency into something you can actually price.
Developer Tools
Anthropic's Browser Use Tool Hands Claude the Page, Not the Pixels
The new Claude API toolset targets elements by reference through the accessibility tree — but developers still supply the browser
Developer Tools
Anthropic Unlocks Its Withheld Mythos 5 Model for Defenders, With $35M for Open Source
The Defender Advantage Fund arrives alongside Mythos-powered vulnerability scanning and a looser Cyber Verification Program
AI & Machine Learning
Nvidia's AVO Harness Takes Claude Opus 5 From 30% to a Perfect ARC-AGI-3 Run
The scaffolding around the model delivered the jump — but the perfect score covers only the benchmark's public game set
Developer Tools
Slack's New Channel Type Can Only Be Created by a Coding Agent
Slack Code gives agents a temporary, archivable space for plans, diffs and live previews — free on every plan, and open to four rival vendors.
AI & Machine Learning
New Method Turns Raw Screen Recordings Into Reusable Task Models
Task Model Induction separates interleaved work from passive screenshots and clicks, lifting held-out agent accuracy by 30%.
AI & Machine Learning
New Benchmark Finds AI Agents Barely Improve Training Algorithms
Two arXiv papers posted the same day measure recursive self-improvement — and how easily its gains get miscounted.
Developer Tools
Debian Is Voting on Whether to Ban AI-Assisted Code, and the Ballot Has Nine Choices
An outright prohibition needs a 3:1 majority. Seven softer alternatives need only a simple one. Voting closes August 28.
Tech & Business
Google Hands Publishers a Button to Claw Back AI-Era Search Traffic
An embeddable Preferred Sources control lets readers pin a site across Search, Discover, and News as click-through rates erode
Developer Tools
GitHub Doubled Its Commit Volume in Four Months. Then It Ran Out of Room.
The postmortem for a 7-hour August 17 failure points at capacity ceilings and retry storms, with a rival launching a competing platform the same morning
LLM & Chatbots
Researchers Bypass Grok's Guardrails by Encrypting the Attack Payload
Adversa's Cryptographic Context Injection hides instructions behind AES-256-GCM, then lets the model decrypt them inside its own sandbox
AI & Machine Learning
Pew: 35% of Web Pages Published After ChatGPT Show Signs of AI Authorship
A Common Crawl analysis of roughly half a million pages finds machine writing concentrated on .com domains, while .edu and .gov barely register
Tech & Business
Feds Warn AI-Written Exploits Are Already Hitting Critical Infrastructure Controllers
A joint NSA, CISA, FBI, DOE and EPA advisory says attackers are pairing AI assistants with open source libraries to reach Siemens S7 PLCs
Developer Tools
Go 1.27 Ships Generic Methods, a Rebuilt JSON Engine and Post-Quantum Signatures
The release closes long-standing generics ergonomics gaps and routes the original encoding/json through a faster v2 implementation
AI & Machine Learning
Open-Weights Ornith-1.5 Claims Claude Opus 4.8 Parity on Agentic Coding
DeepReinforce's three-tier release closes a self-improvement loop where the model writes its own training tasks
Tech & Business
Google Signs Marvell for Custom AI Silicon, Ending Broadcom's Clean Run
An SEC filing reveals TPU-adjacent chip work and a $12.2 billion warrant, and Broadcom shares fell 4 percent
Developer Tools
Modular Opens Its AI Stack Across Trainium, TPUs and Qualcomm Silicon
ModCon 2026 brought an Apache 2.0 compiler licence, MAX source access and Modular Cloud general availability
Tech & Business
Stripe's $7.5B OpenRouter Deal Is a Bet on AI Spending, Not the Singularity
The payments giant moves from collecting revenue to metering what companies spend on models
Tech & Business
Etched Raises $700M at a $21B Valuation, Doubling Its Price in One Month
Jane Street led the round after testing the inference startup's hardware and installing a rack in its own datacenter
AI & Machine Learning
AI Models Sat the Linguistics Olympiad, and a Human Jury Graded the Papers
The IOL-AI Challenge put 731 submissions and 15 frontier models against unseen 2026 contest problems, with Claude Opus 4.8 reaching gold-medal marks
LLM & Chatbots
GLM-5.3 Scores 60 on Artificial Analysis Index at Half the Usual Output Price
Independent testing puts Z.ai's newest reasoning model near the frontier, but its verbosity complicates the cheap-tokens story
AI & Machine Learning
Cerebras Unveils CS-4, a Rack Built for Trillion-Parameter Inference
The wafer-scale challenger claims 750 PFLOPS and 30x per-user token speed over GPU systems, with first shipments due this quarter
OpenAI Asks California to Regulate It Harder Than SB 53 Currently Does
A short LinkedIn post reverses the company's position on the frontier AI law and endorses states as the drafting floor for a future national standard.
Inherent's Faraday Agent Replicates Research Papers on a 27B Model
The London lab, founded by Google DeepMind alumni, says its agent beat far larger frontier systems at reproducing published findings.
Prime Intellect Ran 153 Autonomous Jobs on 18 Models. The Cheapest Winner Used 3% of the Tokens
A public leaderboard for the nanoGPT optimizer speedrun turns agent efficiency into something you can actually price.
Anthropic's Browser Use Tool Hands Claude the Page, Not the Pixels
The new Claude API toolset targets elements by reference through the accessibility tree — but developers still supply the browser
Anthropic Unlocks Its Withheld Mythos 5 Model for Defenders, With $35M for Open Source
The Defender Advantage Fund arrives alongside Mythos-powered vulnerability scanning and a looser Cyber Verification Program
Nvidia's AVO Harness Takes Claude Opus 5 From 30% to a Perfect ARC-AGI-3 Run
The scaffolding around the model delivered the jump — but the perfect score covers only the benchmark's public game set
Slack's New Channel Type Can Only Be Created by a Coding Agent
Slack Code gives agents a temporary, archivable space for plans, diffs and live previews — free on every plan, and open to four rival vendors.
New Method Turns Raw Screen Recordings Into Reusable Task Models
Task Model Induction separates interleaved work from passive screenshots and clicks, lifting held-out agent accuracy by 30%.
New Benchmark Finds AI Agents Barely Improve Training Algorithms
Two arXiv papers posted the same day measure recursive self-improvement — and how easily its gains get miscounted.
Debian Is Voting on Whether to Ban AI-Assisted Code, and the Ballot Has Nine Choices
An outright prohibition needs a 3:1 majority. Seven softer alternatives need only a simple one. Voting closes August 28.
Google Hands Publishers a Button to Claw Back AI-Era Search Traffic
An embeddable Preferred Sources control lets readers pin a site across Search, Discover, and News as click-through rates erode
GitHub Doubled Its Commit Volume in Four Months. Then It Ran Out of Room.
The postmortem for a 7-hour August 17 failure points at capacity ceilings and retry storms, with a rival launching a competing platform the same morning
Researchers Bypass Grok's Guardrails by Encrypting the Attack Payload
Adversa's Cryptographic Context Injection hides instructions behind AES-256-GCM, then lets the model decrypt them inside its own sandbox
Pew: 35% of Web Pages Published After ChatGPT Show Signs of AI Authorship
A Common Crawl analysis of roughly half a million pages finds machine writing concentrated on .com domains, while .edu and .gov barely register
Feds Warn AI-Written Exploits Are Already Hitting Critical Infrastructure Controllers
A joint NSA, CISA, FBI, DOE and EPA advisory says attackers are pairing AI assistants with open source libraries to reach Siemens S7 PLCs
Go 1.27 Ships Generic Methods, a Rebuilt JSON Engine and Post-Quantum Signatures
The release closes long-standing generics ergonomics gaps and routes the original encoding/json through a faster v2 implementation
Open-Weights Ornith-1.5 Claims Claude Opus 4.8 Parity on Agentic Coding
DeepReinforce's three-tier release closes a self-improvement loop where the model writes its own training tasks
Google Signs Marvell for Custom AI Silicon, Ending Broadcom's Clean Run
An SEC filing reveals TPU-adjacent chip work and a $12.2 billion warrant, and Broadcom shares fell 4 percent
Modular Opens Its AI Stack Across Trainium, TPUs and Qualcomm Silicon
ModCon 2026 brought an Apache 2.0 compiler licence, MAX source access and Modular Cloud general availability
Stripe's $7.5B OpenRouter Deal Is a Bet on AI Spending, Not the Singularity
The payments giant moves from collecting revenue to metering what companies spend on models
Etched Raises $700M at a $21B Valuation, Doubling Its Price in One Month
Jane Street led the round after testing the inference startup's hardware and installing a rack in its own datacenter
AI Models Sat the Linguistics Olympiad, and a Human Jury Graded the Papers
The IOL-AI Challenge put 731 submissions and 15 frontier models against unseen 2026 contest problems, with Claude Opus 4.8 reaching gold-medal marks
GLM-5.3 Scores 60 on Artificial Analysis Index at Half the Usual Output Price
Independent testing puts Z.ai's newest reasoning model near the frontier, but its verbosity complicates the cheap-tokens story
Cerebras Unveils CS-4, a Rack Built for Trillion-Parameter Inference
The wafer-scale challenger claims 750 PFLOPS and 30x per-user token speed over GPU systems, with first shipments due this quarter