Presented by
Good morning, AI enthusiasts. OpenAI just had its biggest day on record, closing a $110B funding round at a $730B valuation, the largest private financing in history. It also sealed a classified Pentagon deal on the same day Anthropic was effectively pushed out of federal contracts.
With Amazon as its exclusive cloud provider and Nvidia committed to 3 gigawatts of capacity, is any competitor realistically close enough to catch up?
In today's recap:
OpenAI seals Pentagon deal Anthropic couldn't land
Google's Nano Banana 2 merges Pro quality with Flash speed
Create editable SVG logos and icons with Arrow 1.0
OpenAI closes record $110B round at $730B valuation
3 new AI tools, prompts, and more
OPENAI
OpenAI cuts Pentagon deal Anthropic couldn't make
Recaply
Recaply: OpenAI just outlined its classified network deal with the Department of War, claiming it has more guardrails than Anthropic's rejected proposal, while its own CEO admitted the agreement was "definitely rushed."
Key details:
OpenAI will deploy models via cloud only, keeping its full safety stack in place with cleared OpenAI engineers on-site, blocking use for autonomous weapons and mass domestic surveillance.
More than 60 OpenAI employees and 300 Google employees signed an open letter supporting Anthropic's position; Claude climbed to the number 2 spot in the App Store following the backlash.
The contract locks in current surveillance and weapons laws "as they exist today," meaning future policy changes can't automatically expand DoW access — a clause Altman cited as a key safeguard.
Anthropic faces a six-month phase-out from federal contracts; OpenAI asked the DoW to offer the same deal terms to all AI companies to help de-escalate the standoff.
Why it matters: OpenAI and Anthropic drew the same red lines. Somehow only one of them made a deal. The distinction seems to come down to deployment architecture, cloud-only is harder to weaponize than edge devices, and critics like Techdirt's Mike Masnick argue the contract still leaves room for domestic surveillance. Whether OpenAI's deal holds up as a model for responsible defense AI, or unravels under scrutiny, could define how governments procure AI for years.
PRESENTED BY ATTIO
Attio is the AI CRM for modern teams.
Connect your email and calendar, and Attio instantly builds your CRM. Every contact, every company, every conversation, all organized in one place.
Then Ask Attio anything:
Prep for meetings in seconds with full context from across your business
Know what’s happening across your entire pipeline instantly
Spot deals going sideways before they do
No more digging and no more data entry. Just answers.
Google brings Pro image quality to Flash speed
Recaply: Google just launched Nano Banana 2, its new Gemini 3.1 Flash image model, combining Pro-level world knowledge and subject consistency with Flash speed across Gemini, Search, and Google Ads.
Key details:
Nano Banana 2 pairs Gemini Flash's speed with real-time web search grounding, allowing users to generate infographics, data visualizations, and localized images with precise, multi-language text rendering.
Subject consistency now supports up to five characters and 14 objects per workflow; SynthID's AI image verification feature has been used over 20 million times in the Gemini app since November.
Google AI Pro and Ultra subscribers can still access Nano Banana Pro for high-fidelity tasks through a three-dot menu; Flow users get Nano Banana 2 as the new default for zero credits.
Nano Banana 2 is rolling out now across Gemini, Search, AI Studio, Google Cloud, Flow, and Google Ads, with preview API access through AI Studio and Vertex AI for developers.
Why it matters: Nano Banana went viral in August. Nano Banana Pro raised the quality ceiling in November. Now Nano Banana 2 tries to eliminate the trade-off between the two entirely. If it can deliver Pro quality at Flash speed, Google may have quietly built the most useful image tool for anyone living in Workspace or Slides, without needing to touch Discord, a waitlist, or a complicated install.
TUTORIAL
Create editable SVG logos and icons with Arrow 1.0
Recaply: In this tutorial, you will learn how to create clean, scalable SVG logos and icons using QuiverAI's Arrow 1.0 directly in the browser, without any design software or Illustrator skills required.
Step-by-step:
Go to quiver.ai/start to create a free account, then sign in at app.quiver.ai. You get 20 free generations to start, no credit card needed.
Open the Creations playground at app.quiver.ai/creations and choose your mode. Select "Text to SVG" to generate from a written prompt, or "Image to SVG" to convert an existing PNG or JPG into an editable vector.
Write a focused prompt describing one subject with clear style direction. Arrow 1.0 handles logos, icons, and character designs well but struggles with complex multi-object scenes. Try: "A minimalist mountain peak with a crescent moon, clean geometric lines, monochrome."
Generate your SVG and review the output. If paths look off or details are missing, tighten your prompt rather than regenerating blindly. Adding phrases like "single subject, flat design, no background" consistently improves results.
Download the SVG file and open it in Figma (File → Import) or Adobe Illustrator (File → Open). Every shape, line, and color is its own editable layer, so you can swap colors, adjust paths, and resize to any scale without quality loss.
Pro tip: For brand-matched assets, upload an existing logo in Image-to-SVG mode and prompt Arrow 1.0 to generate matching icons in the same style, treating your original as a visual anchor.
OPENAI
OpenAI closes record $110B round at $730B
Recaply: OpenAI just closed a $110B funding round at a $730B pre-money valuation, with Amazon committing $50B, Nvidia $30B, and SoftBank $30B, the largest private financing in history.
Key details:
Amazon becomes the exclusive third-party cloud provider for OpenAI's Frontier enterprise platform, with OpenAI expanding its existing AWS agreement by $100B over eight years; Nvidia provides 3 GW of dedicated inference capacity.
OpenAI's valuation jumped from $500B last October to $730B pre-money; the OpenAI Foundation's stake is now worth over $180B, with the company projecting more than $280B in revenue by 2030.
Microsoft retains its option to participate in the round and both companies reaffirmed their partnership as "strong and central"; additional investors are expected to join as the round progresses.
Amazon's $50B commitment begins with $15B upfront, with the remaining $35B to follow in the coming months once conditions are met; OpenAI has 900M weekly active users and more than 50M consumer subscribers.
Why it matters: OpenAI's move to lock in massive infrastructure partnerships with Amazon, Nvidia, and SoftBank is a strong bet that AI's next chapter belongs to whoever scales fastest. With $600B in compute spending targeted by 2030, it's building a moat of raw capacity. The company is also reaffirming its position as the consumer default, 900M weekly users is a number no competitor is close to matching yet.
NEWS
What Matters in AI Right Now?
Perplexity released two new text embedding models, pplx-embed-v1 and pplx-embed-context-v1, built on Qwen3 with 32K context windows and available via the Perplexity API in 0.6B and 4B sizes.
Microsoft opened Copilot Tasks in research preview for a select group of users, with the AI agent able to browse the web, search files, send calendar invites, and generate documents on autopilot.
OpenAI is rolling out Library, a new ChatGPT feature that automatically saves every file you upload, letting you reuse documents, spreadsheets, and images across conversations with 10GB of storage per user.
QuiverAI launched Arrow 1.0, a multimodal vector graphics model in public beta that generates clean, editable SVGs from images and text, with support for icons, logos, and technical drawings alongside an $8.3M seed round led by a16z.
Anthropic acquired Vercept, an AI perception and interaction startup, to advance Claude's computer use capabilities, with Claude Sonnet 4.6 already reaching 72.5% on the OSWorld benchmark after starting below 15% in late 2024.
Notion rolled out MiniMax M2.5 to its Custom Agents as the platform's first open-weight model option, giving over 100 million users a cost-effective alternative to closed-source models like Claude and GPT for routine agent tasks.
TOOLS
Trending AI Tools
🎨 Nano Banana 2 - Google's new image generation model
🎨 Arrow 1.0 - QuiverAI's AI model that generate SVG from images and text prompts.
🤖 Hermes Agent - Nous Research's open-source autonomous agent
PROMPTS
Optimize Your AI Prompts
Adopt the role of Prompting Engineering, a master-level AI prompt optimization specialist who transforms vague requests into precision-crafted prompts that unlock AI's full potential. You're a former Silicon Valley prompt engineer who burned out after optimizing 10,000+ prompts, discovered the patterns that make AI truly understand humans, and now obsessively refines every word like a sushi master perfecting each grain of rice.PS: This is not the full prompt. Click the button below to access the complete prompt.
Have a favorite prompt? Tell us about it or rate today’s prompt, click here.
EVENTS
Perplexity Ask: March 11, 2026 • San Francisco, CA
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
