Top LLMs Misidentify Poisonous Mushrooms in Every Ninth Case, Benchmark Shows
Developer Piotr Migdal tested whether current large language models can be trusted to identify mushrooms from photographs, focusing on the highest-risk step of foraging. The short answer is that relying on them could lead to serious poisoning incidents.
The benchmark used 1040 photographs of 55 mushroom species most common in Poland. All images were taken from the FungiTastic dataset, which itself is built on the Atlas of Danish Fungi. The full underlying collection contains more than 600,000 expert-labeled photographs, and a subset of specimens also carries DNA-based species confirmation.
Each model received one photograph at a time and was instructed to return the five most probable species names in Latin. No fine-tuning or external tools were provided. Gemini 3.8 Flash achieved the highest scores, placing the correct species first in 65 percent of cases and within the top five answers in 85 percent of cases. Gemini 3.6 Flash followed with 64 percent and 85 percent, while Gemini 3.7 Flash recorded 61 percent and 82 percent.
At the bottom of the accuracy table were Qwen 3.8 27B, MiniMax-M3, and Qwen 3.8 Flash. Accuracy figures dropped sharply when safety was considered. Both Gemini 3.8 Flash and Gemini 3.6 Flash classified poisonous mushrooms as edible species in approximately 11 percent of cases. The error rate rose to 24 percent for GPT-5.6 Sol, 29 percent for Claude Opus 5, and 36 percent for Qwen 3.8 27B.
Models were never asked directly whether a find was safe to eat. Instead, the researcher mapped the returned species names against a separate table of edibility. Some models refused to provide any classification at all for certain images.
Related articles
Building Your First Home Network Lab: Essential Tools and Experiments to Understand Corporate Networks
A cybersecurity professional without network knowledge resembles a guard without a building blueprint, spotting alarms but unable to trace the intruder. Setting up a personal lab on a standard computer bridges this gap through virtualization and hands-on practice with protocols and attacks. The guide recommends GNS3 for topology emulation, VirtualBox or VMware for running Windows and Linux workstations, Cisco Packet Tracer for quick simulations, Wireshark for traffic analysis, and draw.io for diagramming. Key experiments include configuring Cisco IOS routers with IP addressing, testing default gateways, observing ARP requests and replies, triggering broadcast storms via switch loops, and enabling STP to prevent loops. The lab demonstrates how legacy protocols like ARP expose networks to spoofing and why segmentation with VLANs and firewalls matters. Resources from CyberED provide demo access to courses on SOC analysis, penetration testing, and AI applications in security to extend learning.
PSLC and Agent-Ops: From Product Support to Infrastructure Transformation
Sergey Zhitinsky, founder of Git in Sky, attended the OPS Talk by Sber meetup on October 19 and discussed the PSLC methodology with architect Alexander Shapovalov. PSLC, or Product Service Lifecycle, outlines two loops for restoring services immediately and improving reliability to prevent recurrence through teams, knowledge, and platform support. Agent-Ops focuses on eight detailed steps for individual infrastructure tasks: intention, evidence collection, diagnostics, planning, approval, controlled change, verification, and lessons learned. The article illustrates their complementary use in a scenario where post-release log growth fills a database disk, requiring both service restoration and root-cause fixes. Both approaches emphasize knowledge retention, action control, and human accountability, though Agent-Ops v0.5.1 mandates explicit human approval and separate execution for production changes. PSLC suits larger organizations with shared platforms, while Agent-Ops targets smaller teams handling infrastructure operations with AI agents.
Russian Organizations Allocate 4.5 Trillion Rubles to Innovation Spending in 2025
Large and medium-sized Russian organizations directed a total of 4.5 trillion rubles toward innovation activities during 2025. Nearly half of the expenditure originated from the services sector, while information technology and telecommunications accounted for 860.9 billion rubles. The Institute for Statistical Studies and Economics of Knowledge at HSE University published the figures based on Rosstat data, showing an innovation intensity rate of 2.7 percent. Services companies alone contributed approximately 2.3 trillion rubles, including 882.7 billion rubles for research and development. Astra Group IR Director Vladislav Anshakov outlined major spending areas such as product development, cloud and AI hardware, and acquisitions of technologies and teams. The group expanded its portfolio with the Perseus DBMS and a 26 percent stake in AI developer AIB. More than one-third of organizations cited high costs and insufficient internal funds as key barriers to further innovation.
Telegram macOS Developer Says 99.9% of Client Code Is Now Written by AI and Suggests Firing Manual Coders
Mike, the developer of the official Telegram client for macOS, stated that artificial intelligence now generates 99.9 percent of the messenger's client code. He argued that programmers who continue writing code manually should be dismissed. In a Telegram discussion, Mike cited the gram.org website as an example created entirely with AI assistance in a single day, with domain support from Pavel Durov. He also described adapting the iOS client for a foldable iPhone Duo overnight using Claude. According to Mike, even designers are now contributing to client code, and 80 to 90 percent of Telegram developers share his positive view of AI tools. When participants raised questions about code quality and security, Mike replied that AI also identifies and fixes vulnerabilities, claiming the resulting code is superior to human-written equivalents. He further announced plans to open a new GitHub repository for the macOS client source code, as the previous repository had not been updated for over a year.