Ai-Weekly on Marcin Kucharski

AI Weekly #7: When Investment Velocity Meets Governance Collapse

Fri, 13 Feb 2026 00:00:00 +0000

TL;DR: CISA director uploaded classified documents to ChatGPT. UMG filed $3B lawsuit against Anthropic for illegally training on copyrighted music. Meta committing $115-135B capex in 2026. Governance hasn’t scaled with velocity. Indemnification clauses are now business-critical.

Madhu Gottumukkala, acting director of CISA, uploaded sensitive government documents to public ChatGPT in August 2025.

Universal Music Group filed a $3 billion lawsuit against Anthropic over 20,000+ songs, escalating 40x from their October 2023 suit that covered 500 songs and $75 million in damages. The complaint names CEO Dario Amodei personally and claims Anthropic torrented pirated tracks from LibGen to avoid licensing fees.

AI Weekly #6: 85% Customize Agents, 25% Ship Them

Fri, 23 Jan 2026 00:00:00 +0000

TL;DR: 85% customize agents, 25% ship them. The gap isn’t technology—it’s governance. Security exploits (Gemini Calendar, Cursor IDE, Claude Code) blocked deployment plans in January 2026. The solution: instrument agents like microservices (72% of enterprises now do this), red team for prompt injection, and prioritize vertical agents for defensive use cases (ITOps 44% ROI, cybersecurity 27%) that ship faster than generative ones.

85% of enterprises are customizing AI agents for their business needs.

AI Weekly #5: $10B Infrastructure vs 95% ROI Failure

Fri, 16 Jan 2026 00:00:00 +0000

TL;DR: $37 billion spent, 95% zero ROI. The paradox: infrastructure booming ($10B Cerebras deal, $4.75B power acquisition), but enterprise execution collapsing. Winners have data maturity, cross-functional alignment, and measurable use cases. Losers build custom agents, lose 19% efficiency on complex tasks, and stay stuck in pilot purgatory. 2026 is the year CFOs demand proof.

95% of companies are getting zero return on their AI investments.

Let that sink in. Despite $37 billion spent on GenAI in 2025 alone, MIT’s “GenAI Divide” report shows the vast majority of organizations have nothing to show for it—no measurable ROI, no impact on P&L, just expensive pilots gathering dust.

AI Weekly #4: Your 2026 AI Roadmap Just Hit a Physical Wall

Mon, 29 Dec 2025 00:00:00 +0000

TL;DR: $27 billion in capital moves this week signal AI infrastructure consolidation. Nvidia controls training+inference via Groq acquisition. ServiceNow builds agent control layers. Off-boarding procedures become security-critical when AI accelerates insider threats. Infrastructure constraints, not software, now dominate 2026 roadmaps.

The Reality Check

I spent the last week digging through infrastructure and security reports, and honestly, it forced me to rewrite my entire perspective on 2026.

As tech leaders, we usually live in the software layer: we talk about prompts, agents, and custom workflows. But recent moves by the industry giants made one thing painfully clear: the software party is now being dictated by the people pouring the concrete and buying the silicon.

AI Weekly #3: The Week AI Got Standards

Mon, 15 Dec 2025 00:00:00 +0000

TL;DR: Three major AI standards (MCP, AGENTS.md, Goose framework) finally give enterprises vendor flexibility. Model routing becomes mandatory as pricing diverges 56x between frontier models. Technical implementation now determines competitive advantage.

Three years after ChatGPT launched, the AI industry finally got standards.

OpenAI, Anthropic, Google, AWS, and Microsoft announced the Agentic AI Foundation (AAIF) under Linux Foundation governance this week. The standards: Model Context Protocol (MCP) for data integration, AGENTS.md for project structure, and the goose framework for local execution.

AI Weekly #2: 88% Deployed, Only 6% Ready

Mon, 08 Dec 2025 00:00:00 +0000

TL;DR: 88% of companies deployed AI, but only 6% have infrastructure ready. The real gap: 71% of time spent on manual data integration. Frontier workers (top 5%) are 6-17x more productive due to infrastructure, not talent. Budget shifted 76% toward buying platforms instead of building.

88% of companies adopted AI. Only 6% are ready for it.

Major reports dropped this week: McKinsey says nearly everyone’s using AI, CData says almost nobody has the infrastructure to support it.

AI Weekly #1: The Multi-Provider AI Era Just Started

Fri, 05 Dec 2025 00:00:00 +0000

TL;DR: The AI market entered permanent competition mode. OpenAI’s internal “Code Red,” Gemini 3’s 650M users, and Anthropic’s 300K enterprise customers prove single-vendor stability is gone. Enterprises need multi-provider architectures, abstraction layers, and vendor-independent governance to avoid costly lock-in and migration premiums. The AI giants are already hedging—yours should too.

TOP STORY: The Multi-Provider Era Just Started (And Most CTOs Aren’t Ready)

OpenAI declared internal “Code Red” this week.

Not because of technology failure. Because of competition.