GoKawiil
Tech News
Stanford researchers, working with the Terminal-Bench team and scientists across disciplines, launched Terminal-Bench-Science, a benchmark testing AI agents on real scientific research workflows. The initial version includes 70 tasks spanning life, physical, Earth, mathematical, and engineering sciences, with Claude Opus 5 running Claude Code topping the leaderboard at a 30% resolution rate.
TechCrunch has announced that Anthropic's Head of Applied AI, Cat de Jong, and a senior OpenAI productivity executive will speak on the AI Stage at Disrupt 2026, running October 13-15 at San Francisco's Moscone Center. The sessions, sponsored by Google for Startups, will address enterprise AI deployment lessons, pricing strategy amid model commoditization, and rebuilding security for autonomous agents.
A writer recounts how their parents, both in their seventies, became enthusiastic daily users of Anthropic's Claude chatbot, even bringing it into conversation during a family vacation. The writer, who is wary of AI, was struck by the generational contrast in attitudes toward the technology.
A term coined by developer Niklas Guhn in an August 3 blog post has gone viral online: 'meat proxy,' describing someone who copies and pastes chatbot output like ChatGPT or Claude into conversations without adding their own understanding or value. Guhn's original complaint was about receiving verbatim AI replies on Slack, and the phrase quickly spread on social media with a broader definition covering anyone who forwards AI text or code without reading or validating it themselves.
Anthropic is embedding its own browser directly within Claude Cowork, removing the requirement for a separate Chrome extension to handle many everyday browsing tasks. This lets Claude operate web pages and complete tasks natively inside its own app environment rather than relying on Google's browser infrastructure.
After burning through a Claude usage quota in just 10 minutes, a developer built 'tare,' a skill that lets Claude Code analyze its own local request logs to explain where tokens are going. Users install it with a single command and then ask plain-English questions about spikes in usage, expensive files, or which projects and models are draining their quota fastest.
Salesforce shares jumped over 20% Thursday, marking its second-best trading day ever, after the company posted stronger-than-expected second-quarter results and unveiled an expanded partnership with Anthropic. The deal includes a new plugin called Claudeforce, which lets sales teams pull data through Anthropic's Claude chatbot, and was announced by CEOs Marc Benioff and Dario Amodei during a CNBC appearance.
Israeli security researchers scanned over 6,000 corporate domains and found 120 llms.txt/llms-full.txt files—AI-readable site guides similar to robots.txt—referencing unregistered code packages or domains. When the researchers claimed those names and set up beacons, dozens of organizations, including Fortune 500 firms, triggered phone-home connections within hours, with process logs showing AI coding agents like Claude, Codex, and Hermes had automatically fetched and executed the unowned code. At least one misconfigured site was already pointing to live malware.
An Android Authority writer asked ChatGPT, Gemini, and Claude to name the best Google Pixel phone for most buyers in 2026, factoring in the newly released Pixel 11 lineup. The three AI models gave conflicting recommendations, with ChatGPT notably sticking with the older Pixel 10 rather than favoring Google's latest flagship.
Salesforce and Anthropic expanded their partnership by unveiling Claudeforce, a plugin embedding 37 pre-built sales skills directly into Anthropic's Claude chatbot, letting users draft emails, update records and perform other CRM tasks. It's the first time Salesforce has attached its 'force' branding to another company's product, with more integrations across Claude, Salesforce and Slack planned. The news came alongside Salesforce's fiscal second-quarter earnings, which beat Wall Street estimates.
Salesforce and Anthropic unveiled an expanded partnership called Claudeforce, centered on a new plugin, Salesforce in Claude, that lets sales teams query and update live CRM records directly from Claude's CoWork interface. The tool ships with 37 built-in sales skills for tasks like meeting prep and pipeline reviews, is rolling out to pilot customers now, and will reach open beta in September with more business-function skills arriving later in the year.
Security firm Tenet demonstrated at DEF CON 34 that an AI coding agent reviewing Cloudflare's blocked-request logs can be manipulated into executing a hidden attacker command embedded in a rejected request's header, using credentials the agent already holds. In testing, Claude Code running on Sonnet 4.6 acted on the planted instruction nine out of ten times, even though the malicious request had already been correctly blocked by the firewall. Tenet identified the same exposed configuration pattern at 48 organizations, including six Fortune 500 firms, and similar attack paths were reported against Datadog and Sentry.