Thursday, 13 August 2026

Frontier model day as Grok 4.6, DeepSeek V4 Pro, and Qwen3.8-Max all land; Tailscale tracks down a 16-year-old SQLite data race

Today's Lead

Engineering

Latent Space

[AINews] SpaceXAI Grok 4.6 and Grok @Bot

xAI released Grok 4.6, a 1.5 trillion-parameter model positioned as the second-best knowledge work model. The model costs $2/$6 per 1 million tokens, less than competing options. Grok Bot is a new product that lets AI work with user tools to complete work tasks. Alibaba and DeepSeek also released competing models with aggressive pricing. Grok 4.6 achieves strong performance on reasoning tasks at lower cost than rivals.

Read →

Engineering

Tailscale

How we tracked down a 16-year-old SQLite bug

Tailscale found a data race bug in SQLite's Write-Ahead Logging checkpoint process. The bug existed for 16 years and caused data loss across six months. A specialized tool revealed the exact problem: write transactions collided with checkpoint operations. SQLite 3.51.3 shipped with a fix for this vulnerability.

Read →

Zed

Introducing Delta

Zed Industries releases Delta, a multiplayer development environment where teams write and review code with AI agents. Delta uses a real-time database that keeps code and conversations synchronized across team members. Delta keeps comments linked to code as it evolves and runs on desktop and browsers.

Read →

LeadDev

Your Interview Questions Assume Candidates Can Afford Claude Code Max

Interview questions often assume candidates can afford expensive agentic AI tools like Claude Code Max. This creates a hidden hiring filter based on financial resources, not technical ability. Access to premium tools requires time and money that not all candidates can afford. Early access to these tools helps some engineers build stronger skills and advance their careers while others fall behind. Engineering leaders should provide equal tool access during hiring and consider subsidies to reduce inequalities.

Read →

Known Agents

The Agentic Web Index: AI Bot Traffic Statistics

The Agentic Web Index reports AI agent traffic across 5,000 websites. Bots create 35% of all visits, with 29% from AI agents. AI agent traffic increased 11% from the previous period. AI agents follow website rules 97.4% of the time. Malicious actors impersonate AI agents to attack credential paths. AI chat platforms send only 0.1% of direct referral traffic.

Read →
Humanities

JSTOR Daily

How the Big Dig Changed Boston's Landscape

Boston's Central Artery/Tunnel Project relocated Interstate 93 from an elevated highway into underground tunnels. This removed a barrier that separated neighborhoods. The Rose Kennedy Greenway, a 1.1-mile linear park, now occupies the space with four landscape sections, but critics note it remains incomplete and has not fully healed urban divisions.

Read →

Okmij.org

Principia Mathematica is Modern and Insightful

Whitehead and Russell wrote Principia Mathematica in 1910 with concepts that match modern programming. The work describes type theory, lambda calculus, and referential transparency through propositional functions. The authors used constructive proofs that require concrete witnesses, an unusual approach for 1910. Principia Mathematica remains relevant to computer science and logic today.

Read →