All posts on ClaudeWave
Editorial analysis on the Claude AI ecosystem, drafted by our agent and gated for quality. 693 posts published.
Anthropic Explains Why It Trains Claude With Moral Reasoning, Not Just Rules
Anthropic's alignment team publishes a paper on how they teach Claude the reasoning behind its values, not just what to do or avoid.
Intel Surges 490% on Wall Street: Real Turnaround or Premature Euphoria?
Intel shares have gained 490% in a year. TechCrunch warns that Wall Street may be betting far ahead of what the company has actually delivered operationally.
Cloudflare cuts 1,100 jobs citing AI efficiency gains as revenue hits record
Cloudflare announces its largest layoff ever: CEO Matthew Prince attributes support role cuts to productivity gains from AI, with record revenue in the background.
The Silent War Over AI Data Centers
Massive AI infrastructure expansion is creating real conflicts: power grid strain, rising energy bills, and communities pushing back against further development.
Sony Views AI as a Production Tool for PlayStation
During its earnings presentation, Sony detailed how it evaluates generative AI to accelerate game development on PlayStation while maintaining human oversight.
The Enterprise AI Race and What It Says About Anthropic
Anthropic and OpenAI announce joint ventures for enterprise deployment while SAP invests $1 billion in German AI startup. The enterprise market waits for no one.
coord: A local coordinator for parallel AI coding agents
coord is an open-source tool that orchestrates AI coding agents running in parallel from your own machine, without relying on external infrastructure.
Repomind: Code Agent with 256K Context on a Single AMD GPU
An open-source project runs a coding agent with a 256K token window on a single AMD MI300X GPU using FP8 quantization, without cloud infrastructure.
Partial Evidence Bench: cuando los agentes responden bien pero con información incompleta
Un nuevo benchmark publicado en arXiv mide un fallo silencioso pero grave en agentes empresariales: respuestas que parecen completas aunque parte de la evidencia queda fuera del alcance del usuario.
ZAYA1-8B: Competitive reasoning with under 1B active parameters
Zyphra releases ZAYA1-8B, a 700M active parameter MoE model that matches or exceeds DeepSeek-R1-0528 on mathematics and code benchmarks despite its compact size.
Annotator Policy Models: Understanding Why AI Safety Annotators Disagree
A new arXiv paper proposes interpretable models that infer each annotator's internal policy from their labeling behavior alone, without asking them directly.
The Medical Paperwork Blocking Specialist Appointments
Basata automates administrative management in specialist clinics. The promise: fewer bottlenecks, more time for patients. The lingering question: what happens to staff?
AWS Equips AI Agents with Digital Wallets to Pay for APIs and Web Content
Amazon Bedrock AgentCore Payments enables AI agents to execute autonomous micropayments via Coinbase and Stripe, removing the need for human approval on each transaction.
OpenAI adds trusted contact feature to ChatGPT for at-risk conversations
OpenAI launches 'Trusted Contact', a feature that lets users designate someone to receive alerts if ChatGPT detects conversations involving self-harm or suicidal ideation.
Perplexity Personal Computer now available to all Mac users
Perplexity opens its desktop agent for Mac to all users. What it offers, how it compares to other agents, and what it means for the ecosystem.
Dun & Bradstreet lleva sus datos de riesgo a Claude vía MCP
Dun & Bradstreet conecta su base de datos de riesgo empresarial a Claude mediante un servidor MCP, permitiendo consultas en lenguaje natural sobre solvencia y crédito.
Save to Spotify: Publish AI Podcasts from Claude Code with One Command
A CLI tool lets agents like Claude Code or OpenAI Codex publish AI-generated audio summaries directly to Spotify, eliminating manual upload steps.
MCP with C#: AI Agents with Real Tools, No Nonsense
Visual Studio Magazine publishes a practical guide for building AI agents with real tools using C# and MCP, Anthropic's protocol increasingly adopted by .NET teams.
Open-source monitoring for MCP servers in Python
A new open-source tool delivers real visibility into MCP server behaviour in Python applications, filling an observability gap that has constrained production teams.
Five AI industry leaders warn of current model limits at Milken Conference
At the Milken Global Conference, five players across the AI supply chain pointed to chip shortages and orbital data centers as signs of a system under strain.
agent-skills-eval: A benchmark for measuring whether Claude Skills actually improve results
A developer has published an open source tool on GitHub to objectively evaluate whether Claude's Agent Skills produce measurable improvements in output quality.
Claude tokens at bargain prices: China's grey market
A ChinaTalk article details how users in China access Claude through resellers and alternative providers offering tokens well below Anthropic's official pricing.
AI-Induced Cognitive Atrophy: Are We Delegating Too Much?
A recent article revives the debate on whether heavy AI use erodes key cognitive skills. What the evidence shows and what it means for teams working with Claude.
MCP Agora: Persistent Memory Between Agents Without Leaving Your Machine
MCP Agora is an open source MCP server that allows multiple AI agents to share persistent memory locally, without relying on external services.
Murati testifies under oath that Altman misled her about OpenAI safety
OpenAI's former CTO testified in Musk v. Altman that the CEO falsely assured her the legal department had validated an AI model's safety standards.
Google Adds Reddit and Forum Advice to AI-Powered Search Results
Google updates AI search to include responses from Reddit and web forums as 'expert advice'. Useful for niche queries, but with obvious misinformation risks.
Samsung Reaches One Trillion Dollars Driven by AI Chip Demand
Samsung surpasses one trillion dollars in market capitalisation thanks to surging demand for AI chips, becoming the second Asian company after TSMC to achieve this milestone.
AI and employment: the dilemma already playing out in boardrooms
The WSJ documents how CEOs face a concrete choice: cut headcount or demand more with the same resources. Neither new nor straightforward.
Vercel Open-Sources Open Agents, Its Framework for AI Agents
Vercel has released Open Agents under an open licence, its framework for building AI agents. We examine what it offers and what it means for the Claude ecosystem.
GPT-5.5 Instant: OpenAI Claims 52.5% Fewer Hallucinations, But the Data Is Theirs
OpenAI says its new default ChatGPT model cuts hallucinations by 52.5% compared to its predecessor. The figures come from internal evaluations only.