Three wars are running in parallel right now. One involves missiles near a nuclear reactor. Another is an AI arms race where four companies shipped frontier models in a single week. The third is oil at $106 a barrel dragging every portfolio on earth into the red. Here is the breakdown.
⚡AI
OpenAI merged ChatGPT, Codex, and Atlas into one desktop Superapp with a 1.05M token context window. Google fired back with Stitch, a free design tool that tanked Figma stock 12% in a single session. Anthropic's Claude Opus 4.6 now leads every coding benchmark at 80.8% SWE-Bench. The frontier model war just became a platform war. Pick your stack carefully.
🌎World
Iran fired missiles near Israel's Dimona nuclear center. That is a first. 100+ injured, and Trump gave a 48-hour ultimatum to reopen the Strait of Hormuz. Over 5,000 vessels are stranded. Iran responded by threatening full closure. Diego Garcia, a US-UK base in the Indian Ocean, also took fire. This is not contained anymore.
📈Markets
Everything is red. S&P down 1.51%. NASDAQ down 2.01%. Bitcoin dropped 2.77% to $68,616. Brent crude surged past $106 on Hormuz fears. Goldman says triple-digit oil could last years if the strait stays closed. The sell-off is energy-driven, not tech-driven. Watch oil, not earnings.
💡Wild Card
The Pentagon just made Palantir's Maven a permanent program of record. That means long-term military AI funding across every branch. AI in defense went from experimental to infrastructure overnight. Meanwhile, you can now run DeepSeek and Llama on your phone with zero internet. Private AI is real. The gap between what governments build and what individuals can run is shrinking fast.
Full breakdown below. Start with the infographic.
📊 Daily Intelligence Infographic
Daily Intelligence Infographic
Generated by NotebookLM AI
🎯 Today's Focus
1
Iran-Hormuz Crisis Week 4
Missiles hit near Dimona nuclear center, Trump gives 48hr ultimatum to reopen Strait. Oil above $100/barrel threatening global energy supply.
2
AI Superapp Wars
OpenAI consolidating into desktop Superapp, Google Stitch disrupts design (Figma -12%), Claude Opus 4.6 leads SWE-Bench at 80.8%.
3
Market Stress
S&P down 1.51%, NASDAQ down 2.01%, crypto also falling. Goldman says oil may stay at triple digits for years.
🚀 AI Developments
Superapp
OpenAI Superapp
GPT-5.4 consolidation of ChatGPT, Codex, and Atlas into a single desktop application. Features 1.05M token context window. Represents the largest UX consolidation in AI history, targeting $25B+ revenue.
Tops the intelligence index at 57 points with 1M context window and 77.1% ARC-AGI-2 score. Full multimodal reasoning across text, image, audio, and video makes it the most versatile model released this quarter.
Achieves 80.8% on SWE-Bench Verified, leading all coding benchmarks. Pentagon DOD lawsuit draws support from OpenAI and Google employees. Enterprise deployment expanding via Microsoft partnership.
Free AI design tool that generates high-fidelity reports from natural language prompts. Converts napkin sketches into board-ready executive reports. Figma stock dropped 12% on launch day, signaling a major disruption.
120B-parameter hybrid mixture-of-experts model with only 12B active per forward pass. Built specifically for multi-agent applications. Announced at GTC 2026 as a cornerstone of NVIDIA's AI software stack.
Rivals Claude Opus 4.6 at significantly lower cost, emerging as a new challenger in the frontier model race. Strong performance across coding and reasoning benchmarks positions it as a cost-effective alternative.
Enterprise platform for building custom AI models. Company on track for $1B ARR with CEO confirming enterprise focus is paying off. GPU-efficient architecture differentiates from larger competitors.
Pentagon designates Maven as a formal program of record with long-term military funding across all branches. Marks the transition of AI from experimental defense projects to core military infrastructure.
Users commit at 2x the rate with Plan Mode enabled. Head of Design shows designers coding actual features instead of Figma mockups. Shift+Tab activates structured step-by-step execution for complex tasks.
Run DeepSeek, Llama, Phi, and Gemma on-device with zero internet connectivity. Full privacy-first AI on your phone with no cloud dependency. Works with Siri shortcuts for seamless mobile integration.
Strikes hit Arad and Dimona near Israel's nuclear research center, injuring 100+. First time in history a nuclear facility has been directly targeted. Major escalation in regional conflict with global implications.
Trump gives Iran 48hr ultimatum to reopen strait. Iran says the strait will be completely closed if the US follows through on threats. Over 5,000 vessels stranded as global shipping faces unprecedented disruption.
Iran fired two intermediate-range ballistic missiles at the US-UK military base on Diego Garcia in the Indian Ocean. Neither missile struck the base directly, but the attack marks a significant geographic expansion of hostilities.
Israel launches new strikes targeting Hezbollah infrastructure in southern Lebanon. The operations aim to degrade cross-border attack capabilities amid the broader regional escalation driven by the Iran crisis.
64 killed including 13 children in a hospital strike in western Darfur, confirmed by the WHO. The ongoing conflict continues to produce devastating civilian casualties with limited international intervention.
Cuba's power grid collapsed for the third time in March, leaving the entire country without electricity. The recurring infrastructure failures reflect deep systemic issues with public health and economic consequences.
A suspected meteor lit up skies across Ohio and Pennsylvania accompanied by sonic booms. Residents reported brilliant flashes and shaking windows. Authorities confirmed no damage but investigations continue.
Zain Kahn — “Scientists use plastic to beat Parkinson’s”
Overview
Science Sunday special covering non-AI breakthroughs across space, medicine, and hardware innovation.
Headlines
Hubble catches comet exploding in real time. Man survives 48 hours without lungs. Edinburgh scientists convert plastic bottles to Parkinson's medication. Harvard builds chip to twist and control light.
New Tech
Taya AI necklace for journaling, Nosh One robot chef, iGarden Swim Jet, Fort wearable for strength training.
Good News
Botox shows promise as antidepressant. Wearable compression sleeve aids cancer recovery. New cholesterol pill enlicitide cuts LDL by 60%.
“The odds of witnessing a comet breaking up live through Hubble are vanishingly small”
Pentagon makes Palantir Maven official. White House drops national AI framework. WordPress gives AI agents publish access via MCP integration with 19 write operations.
Google Stitch turns hand-drawn financial sketches into premium executive reports with one prompt. Demonstrates FP&A judgment, not just formatting. One-click deployment from sketch to boardroom-ready output.
Key Insight
The difference between generic AI output and boardroom-ready output is asking for judgment, not design.
Impact
Figma stock dropped 12% on Stitch launch day. This pattern will move into every category of professional work where design meets domain expertise.
“Don’t ask for a design. Ask for FP&A judgment built into a visual format.”
1) Use AI daily — don’t just study it. 2) Schedule quarterly AI opportunity reviews. 3) Try new tools with specific problems in mind, not generic curiosity.
Key Insight
Companies that evaluate AI, find it not ready, and don’t revisit for a year fall behind competitors who were more persistent. Maintain a “not yet but soon” list.
Tools Mentioned
Company Research Agent, Prospect Extractor, ICP Builder on agent.ai.
“The frequency with which we get new [X changes everything] values is accelerating”
GPT-5.4 Superapp consolidation merges ChatGPT, Codex, and Atlas into one desktop app with 1.05M tokens. Targeting $25B revenue. Largest UX consolidation in AI history positions them as the default consumer AI platform.
Gemini 3.1 Pro tops intelligence benchmarks at 57 points with 1M context and 77.1% ARC-AGI-2. Stitch disrupts the design tool category on day one. Strong multimodal reasoning across all modalities gives them unique breadth.
Llama 4 open-source release focuses on agentic capability and developer ecosystem. Strategy: out-distribute rather than out-spend. Building the largest open-source AI community as a competitive moat against closed-source rivals.
On track for $1B ARR with the Forge enterprise platform for custom AI models. GPU-efficient architecture appeals to cost-conscious enterprises. European headquarters provides regulatory advantage as EU AI Act takes effect.
Claude 4.6 leads SWE-Bench Verified at 80.8%, the top coding benchmark score. Pentagon lawsuit draws cross-industry support from OpenAI and Google employees. Enterprise deployment expanding via Microsoft Azure partnership.
Cursor achieves 2x commit rates with Plan Mode. MiniMax M2.5 rivals frontier models at lower cost. Xiaomi MiMo-V2-Pro enters the arena. The challenger tier is compressing the gap with incumbents faster than expected.
Dharmesh's 3-Part Strategy: Use AI daily, schedule quarterly opportunity reviews, adopt a problem-first tool evaluation approach. Don't dismiss AI and forget — revisit regularly with a “not yet but soon” list.
Agentic workflows without code using Gumloop, n8n, or Zapier Central. Simple research agents cost $5-20/month. Four architectures available from single-agent to multi-agent orchestration.
Private LLM runs DeepSeek, Llama, Phi, and Gemma on your phone with zero internet. Full privacy, no cloud dependency. Works with Siri shortcuts for hands-free private AI access.
Google Stitch generates high-fidelity designs from plain language. No wireframes or design background needed. Deploys in one click from sketch to production-ready output across finance, marketing, and operations.
Base44 Superagents run 24/7 — scan inbox, draft replies, follow up with leads, pull weekly reports. No servers or Docker needed. Always-on automation for repetitive business workflows.
Claude Opus 4.6 leads SWE-Bench at 80.8%. Cursor Plan Mode doubles commit rates. WordPress MCP integration adds 19 write operations for AI agents to manage content programmatically.