August 2026 - Page 2 of 52

insurance claims adjusters really hate ai a solid gabled house one chimney

You Know Who Really Hates AI? Insurance Claims Adjusters

Glassdoor’s 2026 workforce study found that 98 percent of AI mentions from insurance claims adjusters are negative, the highest of any occupation, and WIRED’s reporting explains why: automated loss reporting that misroutes claims, hallucinated summaries relayed to claimants, customers who blame the human for the machine, and a job market in which sector employment fell 21 percent in a year and entry-level postings halved. This piece covers the Glassdoor methodology, the labour-market data, what Lemonade and State Farm are automating, the wider turn against AI across the workforce, and what the claims adjuster revolt teaches any organisation deploying AI in front of skilled staff.

Read more
perplexity hybrid mode mac local models subtasks a solid piggy bank

Perplexity to Launch Hybrid Mode on Mac Using Local Models for Subtasks

TestingCatalog has surfaced Hybrid mode, an unreleased setting in Perplexity Computer’s Mac app that keeps orchestration in the cloud while delegating suitable subtasks to a model running locally, plus a Privacy Gate model that checks outbound data for personal information and asks the user before it is sent. This article covers the three downloadable local models (a 19GB Perplexity model, Qwen 32B at 17.4GB, and a 5.6GB Gemma-based model for 16GB Macs), which Mac mini configurations qualify, how Privacy Gate differs from June’s hybrid agentic inference design, the credit arithmetic the feature targets, how it compares with the Nvidia-only Portable Computer, and what a UK business should do about a feature that is still hidden and undated.

Read more
runtime trust ai agents drift expose data memory poisoned a solid ship anchor

AI Agents That Pass Authentication Can Still Drift, Expose Data, or Get Memory-Poisoned

A VentureBeat guest essay argues that AI agents which pass authentication can still drift from their objective, expose data and have their memory poisoned, because authentication verifies who an agent is while runtime trust verifies what it is doing. This piece explains the five runtime threats, the five-pillar runtime trust model (intent validation, behavioural monitoring, policy enforcement, least-privilege execution, human oversight), how it maps onto the six deployment gates in the companion essay, the 2026 research measuring memory-poisoning success rates, the OWASP and MITRE ATLAS entries that name each threat, the survey figures on the post-authentication control gap, what Microsoft, AWS and Cloudflare have shipped, and a roadmap a UK business can run with the tools it already owns.

Read more
ai agents need agent identity before gateway a solid wax seal stamp

AI Agents Need Their Own Identity Before They Need a Gateway

A pair of VentureBeat guest essays argue that AI agents need their own identity before they need a gateway: a proxy can only enforce what it can attribute, and most organisations have no agent identity layer to attribute anything to. This piece explains the borrowed-token problem, the six-gate dependency-gated deployment model that puts the gateway fifth, the LiteLLM CVE-2026-42271 warning, the 2026 survey data on the machine-identity gap (109 machine identities per human, 68% unable to tell agents from people), what Okta, Microsoft, Cloudflare, Kong, the IETF and NIST have shipped, and a 30-day plan a UK business can run without buying anything.

Read more
caterpillar ai deployment mining automation lessons a solid hard hat dome v2

Caterpillar Is Bringing to AI Deployment What It Learned From Automating Mining

Caterpillar has spent more than three decades running autonomous machines at remote mining sites, and CTO Jaime Mineart now says that experience is the template for how the company deploys AI. This piece covers the 8.6 billion tonnes hauled autonomously, the lesson that deploying a machine is not transforming a site, the voice-driven Cat AI Assistant built on 1.6 million connected assets and 16 petabytes of Helios data, AI agents modernising legacy code, the record $20.5 billion quarter driven by data-centre power demand, the $100 million workforce training pledge, and the honest limits of the mining analogy for organisations planning their own AI rollout.

Read more
rote tool task automation ai editors a solid cassette tape

Rote Tool Enables Task Automation in Cursor, Claude, and Multiple AI Editors

Modiqo has wired its Rote capture-and-replay platform into nine AI coding environments, including Cursor, Claude Code, Codex, Windsurf, Kimi and Visual Studio Code, with four releases shipping in the final week of August alone. The Rote tool records what an AI agent actually did during a successful run and compiles it into a deterministic, replayable Play — the vendor’s benchmark puts a replay at roughly 300 tokens against 14,900 for the original discovery run. This piece covers the multi-editor rollout, how capture and replay work, how Plays compare with prompts, skills and workflows, pricing from free to $799 a month, and the limits worth knowing before the Rote Playoffs hackathon stress-tests it all.

Read more
rote playoffs modiqo ai hackathon a solid trophy cup

Modiqo Opens Registration for the Rote Playoffs AI Hackathon

Modiqo has opened registration for the Rote Playoffs, a free, fully online AI hackathon run with the WeMakeDevs community. Entrants capture a repetitive workflow as a reusable “Play” on the Rote platform, publish it, and are judged on whether it runs, whether strangers trust it, and whether others adopt it. Registration closes 1 September 2026, the kickoff livestream is at 4pm UK the same day, and submissions close Sunday 7 September at 8pm London time. First place wins a MacBook Pro; this piece covers the dates, the prizes, the judging, Modiqo’s $3M backstory and how to prepare.

Read more
workbuddy hy4 preview work and coding tasks a solid briefcase

Hy4 preview Is Now Available on WorkBuddy for Work and Coding Tasks

Tencent’s 770B open-weight Hy4 preview is now selectable inside WorkBuddy, the company’s desktop AI agent for office and coding work, and it is free there until 10 September 2026. This piece covers what the integration changes for daily work, how the model switch works and what it is recommended for, the 163-expert blind test that Tencent ran inside WorkBuddy against GLM 5.3 and Kimi K3, the benchmarks that matter for work and coding tasks, the quota and reasoning-time caveats, the API and local GGUF routes for teams that would rather not use the app, and how WorkBuddy compares with Claude Cowork.

Read more
clipdance ai video generator review a solid clapperboard tilted bar

ClipDance Review: Complete Guide to a Powerful AI Video Hub

ClipDance is a browser-based AI video generator that has quietly become one of the more interesting ways to reach the current crop of frontier video models without opening seven different accounts. Rather than building its own model, the platform — which started life as Seedance2.so — sells one credit balance that works across Seedance 2.5, […]

Read more
GLM-5.3-Flash - glm 5 3 flash open weight 320b model a solid lightning bolt

GLM-5.3-Flash: The 320B Open-Weight Model That Ran on Chinese Chips

Z.ai spent six days serving an anonymous model called Ox Alpha on OpenRouter, took nearly 20% of the platform’s weekly token share, and only then revealed it was GLM-5.3-Flash — a 320B mixture-of-experts model with 18B active parameters, a one-million-token context window and MIT-licensed weights. This piece works through the hybrid attention architecture, what the benchmark table supports and what it does not, what the API actually costs once the launch promotion ends, how credible the domestic-silicon claim is, and what any of it changes for a business choosing a model this quarter.

Read more
CHAT