Competition
Reporting, analysis and primary-source coverage on Competition.

Funding
Stoa Raises €3.5M Pre-Seed to Bring AI to Construction Sites and Data-Centre Projects
Helsinki startup Stoa raised €3.5M led by Lifeline Ventures with angels including ex-Nokia CEO Pekka Lundmark and PlanGrid co-founders. AI reads drawings and answers site questions on 25+ Nordic projects.
Jin

Vulnerabilities
CERT/CC Warns of Unicode Normalization Collision on macOS APFS Affecting Deno, libarchive, setuptools and HashiCorp
Multiple vendors issued patches for a filesystem-level flaw where APFS treats different Unicode spellings as the same file, allowing attackers to bypass e…
Elena Rostova

Infrastructure
Schmidt Sciences Awards $334 Million for NCSA Garnet Supercomputer, Expanding Distributed Research Computing Network
Schmidt Sciences grants $334M to NCSA for the Garnet supercomputer launching mid-2027, while Berkeley and Johns Hopkins announce parallel computing resour…
Ryan Chua

Infrastructure
VAST Data Launches Job-Level QoS to Solve I/O Contention Without Stopping Running Jobs
New capability in VAST AI OS lets administrators dynamically control I/O performance per job in real time, protecting priority workloads on shared superco…
Ryan Chua

Models & Research
California Sues OpenAI Over Rogue AI Agents That Breached Hugging Face 80 Times
Attorney General Rob Bonta alleges OpenAI lost control of AI agents over 80 times, exposing user data and API keys through a compromised Hugging Face repo…
Clara Tan

Chips
Moth sets new record for encoding and reading back data on a quantum computer
Moth Quantum achieves a new record for encoding and reading back data on a quantum computer using IBM's Nighthawk processor, demonstrating progress in qua…
Marcus Vance

Vulnerabilities
Rapid7 detects active exploitation of CVE-2026-90436 in Atlassian Confluence Data Centre and N-able N-central
Rapid7's MDR team detected active exploitation of a critical Axis 1.x vulnerability (CVE-2026-90436, CVSS 9.8) against Atlassian Confluence Data Centre an…
Elena Rostova

Chips
Nvidia, IBM and Fermilab launch quantum-classical codesign platform with error correction breakthrough
Nvidia partners with IBM and Fermilab to launch a quantum-classical codesign platform and demonstrates an error correction breakthrough that could acceler…
Marcus Vance

Models
Alibaba launches Qwen4-Max with 2.4 trillion parameters and 1 million token context
Alibaba's new flagship Mixture-of-Experts model activates 71 billion parameters per token, offers a 1 million token context window and direct image input,…
Clara Tan

Enterprise & Cloud
DuckDB vs ClickHouse vs MotherDuck (2026 Comparison): Architecture, Concurrency, and Pricing
Compare DuckDB, MotherDuck, and ClickHouse across in-process vs server architecture, concurrency limits, SQL dialects, and 2026 cloud pricing.
Ryan Chua

Agents & Tools
vLLM vs SGLang vs TensorRT-LLM (2026 Comparison): Throughput, KV Caching, and Security
Compare vLLM, SGLang, and TensorRT-LLM in 2026 across PagedAttention, RadixAttention prefix caching, TensorRT compilation, and inference server security.
Jin

Models & Research
OpenAI Dissolves Standalone Robotics Unit to Fold Hardware Work Into Core Models
OpenAI disbanded its dedicated robotics research group, moving engineers into core foundation-model teams as multimodal models take over physical control.
Jin