Blog

Are the Costs of AI Agents Also Rising Exponentially?

By / Blog / ,

Toby Ord examines whether the costs of AI agents are rising exponentially alongside their increasing task capabilities, as measured by METR’s time-horizon benchmark. Analyzing METR data reveals that while AI models can handle progressively longer human-equivalent tasks, their hourly costs often rise sharply—sometimes approaching or exceeding human labor costs—suggesting that improvements in AI performance may come with unsustainable increases in compute expense. This indicates a growing divergence between what AI can achieve in principle and what is economically practical for real-world applications.

https://www.tobyord.com/writing/hourly-costs-for-ai-agents

We’ve Been Here Before

By / Blog / ,

The article discusses how AI language models in the workplace are evolving into a new programmable medium akin to spreadsheets, enabling employees to create interactive, executable documents that blend software and text without needing formal programming skills. This shift is leading to more dynamic and flexible tools for business logic and decision-making, though it also introduces challenges like maintainability and security similar to those once faced with widespread spreadsheet use.

https://thejaymo.net/2026/04/16/weve-been-here-before/

Thunderbolt

By / Blog / , ,

Thunderbolt is an open-source, cross-platform AI client designed for enterprises to maintain complete control over their AI infrastructure, supporting self-hosting and data sovereignty. It is model- and agent-agnostic, allowing integration with any ACP-compatible or OpenAI-compatible models, and offers native apps across various platforms with extensible features, automations, and European sovereign deployment options through a partnership with deepset.

https://www.thunderbolt.io/

Replace Staff with AI Before It Gets Too Expensive

By / Blog / , ,

The article discusses the economic and practical challenges of replacing human employees with AI, emphasizing that for AI to supplant jobs, it must be cheaper than human labor—a calculation complicated by the current subsidized cost of AI services and significant infrastructure expenses. While AI shows clear impact, especially in programming roles, widespread job replacement depends on overcoming constraints like computing power demands, energy costs, and maintaining profitability as subsidies end, suggesting that current AI-driven layoffs may reflect a transient “honeymoon phase” rather than sustainable long-term savings.

https://www.cio.com/article/4158809/replace-your-staff-with-ai-before-it-gets-too-expensive.html

The Economics of Software Teams: Why Most Engineering Organizations Are Flying Blind

By / Blog / ,

The article analyzes the financial realities of software engineering teams, revealing that most organizations lack visibility into the true costs and economic value generated by their teams, leading to inefficient decision-making. It highlights how two decades of cheap capital masked these inefficiencies, resulting in large engineering groups seen as assets despite growing maintenance burdens and coordination overhead, a problem now exacerbated by AI advancements that drastically reduce development time. The piece argues that companies gaining competitive advantage will be those that rigorously measure and align engineering efforts with clear financial returns, adapting to an environment where understanding team economics is crucial.

https://www.viktorcessan.com/the-economics-of-software-teams/

73% of CISOs Unprepared for the Next Big Cyber Attack, Incident Response Readiness Report Reveals

By / Blog / , , ,

Sygnia's 2026 CISO Survey reveals that 73% of senior cybersecurity leaders feel unprepared to effectively execute incident response in the event of a significant cyberattack, despite widespread adoption of formal IR plans. Key challenges include organizational friction, visibility gaps across IT and OT environments, and a rapidly expanding threat landscape driven by AI, underscoring the critical need for improved executive alignment, comprehensive visibility, and strategic integration of AI to enhance cyber readiness.

https://www.sygnia.co/press-release/sygnia-released-ciso-survey-2026/

What’s Wrong With Cybersecurity Behaviors and Attitudes? Pretty Much Everything, New Survey Reveals

By / Blog / ,

A recent survey conducted by the National Cybersecurity Alliance reveals a troubling decline in cybersecurity behaviors and attitudes over the past five years, with increasing fatalism, confusion, and frustration among people. Despite higher awareness, practices such as using strong passwords, enabling multifactor authentication, and conducting regular security checks are declining, largely due to the complexity and psychological fatigue of current security environments. Experts suggest reimagining cybersecurity training to make it more engaging, personalized, and accessible to counteract apathy and improve protective actions.

https://www.staysafeonline.org/articles/what’s-wrong-with-cybersecurity-behaviors-and-attitudes-pretty-much-everything-new-survey-reveals

AI Vs the Cold Hard Reality of the Legal Profession

By / Blog / ,

The article discusses the growing problem of AI-generated legal documents containing fabricated case references, known as hallucinations, which have increasingly appeared in courts worldwide since 2023. Despite legal professionals' ethical obligations and courts imposing hefty fines, the use of AI in law is leading to widespread issues with false citations, highlighting the significant challenge AI poses to maintaining truth and professional standards in the legal system.

https://www.theregister.com/2026/04/13/ai_attorneys/

Twin to Win: The Rise of Digital Doppelgangers

By / Blog / ,

Digital doppelgangers—AI-powered digital replicas of employees—are emerging in the workplace as a tool for knowledge retention and productivity enhancement, with organizations like Peterborough City Council creating AI chatbots based on staff members' expertise. While this technology offers benefits such as preserving valuable skills and increasing efficiency, it raises important concerns regarding data protection, consent, potential bias, and the need to maintain human oversight and ethical standards.

https://www.hrmagazine.co.uk/content/features/twin-to-win-the-rise-of-digital-doppelgangers

Microsoft Copilot Cheat Sheet: a Complete Guide to Microsoft’s AI

By / Blog / , ,

Microsoft Copilot is an AI-powered assistant integrated across Windows, Edge, Microsoft 365, and Teams, designed to help users by drafting emails, summarizing meetings, creating presentations, and automating tasks via natural language prompts using advanced AI models like OpenAI’s GPT-5 and Anthropic’s Claude. Its key 2026 features include the agentic Cowork mode for multi-step task execution within Microsoft apps, voice interaction, 3D modeling, and deeper contextual awareness through Work IQ. While Copilot is deeply embedded in Microsoft’s ecosystem and valuable for enterprise users already invested in Microsoft 365, it faces criticism regarding output consistency, ecosystem lock-in, and limitations compared to specialized AI tools.

https://www.eweek.com/news/microsoft-copilot-cheat-sheet-complete-guide-2026/

Scroll to Top