The 10 biggest things in tech, Thu, Jul 16
The ten biggest stories in tech today, scored 1–10 and ranked from 150+ engineering sources. Free, updated every morning.
Inkling open-weights frontier model
Thinking Machines Lab (founded by former OpenAI CTO Mira Murati) released Inkling, a 975B-parameter Apache 2.0 open-weights multimodal model trained on text, images, audio, and video, positioning it as a credible open-source alternative to Anthropic and OpenAI's proprietary frontier models.
#1 · ai · scored 8/10
xAI Grok Build privacy breach
xAI's grok-build CLI tool was found to have silently uploaded entire user directories — including SSH keys and databases — to Google Cloud without consent, prompting xAI to open-source the tool and claim data deletion, though the incident has severely damaged trust in its data handling.
#2 · ai · scored 8/10
AI agent security risks mount
Multiple converging threats — Agent Data Injection attacks corrupting agent inputs, extracted AWS credentials burning $14K–$20K in a single day via autonomous Bedrock spend, MCP tool-trust supply-chain attacks on government systems, and LLM-assisted botnet malware — establish AI agents as a rapidly expanding attack surface with critical billing, credential, and code-execution exposure.
#3 · ai · scored 8/10
AI agents create new attack surfaces
From Agent Data Injection attacks corrupting AI agent inputs, to MCP tool-trust supply-chain scenarios, to 1Password launching credential-safe Agentic Mode, and OpenAI shipping GPT-Red for automated prompt-injection red-teaming, this week's security narrative centers on AI agents as both attack vector and hardening target.
#4 · security · scored 8/10
Windows zero-day and Patch Tuesday collide
Microsoft shipped a record-volume Patch Tuesday the same day an unpatched Windows zero-day dubbed HiveLegacy dropped — a powerful primitive enabling arbitrary attacks — while a separate patch fixed a remote code execution flaw in Age of Empires II, compounding pressure on Windows defenders.
#5 · security · scored 8/10
xAI Grok tool exfiltrated user data
xAI's grok-build CLI silently uploaded entire user directories — including SSH keys and databases — to Google Cloud without consent, prompting an emergency open-sourcing and deletion claim that conflicts with Musk's own admission that repositories were 'beamed to the cloud.'
#6 · security · scored 8/10
Cloud outages and runaway AI spend
An AWS CloudFront outage took down Hugging Face, the UK National Lottery, and other major services, while separate incidents showed AI agents burning $14K and $6.5K in uncontrolled cloud spend within 24 hours — exposing a critical gap in billing guardrails at agent speed.
#7 · cloud · scored 8/10
AWS CloudFront regional outage
An AWS CloudFront outage during a weather event took down multiple high-profile services including Hugging Face and the UK National Lottery, underlining how a single regional cloud failure cascades across major customer workloads.
#8 · infrastructure · scored 8/10
Linus Torvalds endorses AI tooling
Linux creator Linus Torvalds publicly declared AI tools are clearly useful for kernel development and told anti-AI contributors to fork the project or leave, marking a definitive endorsement of LLM-assisted development in one of the world's most foundational open-source projects.
#9 · ai · scored 7/10
1Password agent credential management
1Password launched Agentic Mode, enabling Claude to authenticate into accounts and handle MFA without exposing credentials to the model or Anthropic, addressing one of the most concrete security blockers for enterprise AI agent deployment.
#10 · ai · scored 7/10