Issue 002 · Friday, May 22, 2026
About this Notebook
I'm building toward a for-profit AI venture that funds Gospel work. Each Friday I share the news, tools, and patterns I'm watching, with builders in mind. Join me as we figure out what kind of company God might use to change how Gospel initiatives are funded.
— Tommy Lee, by Friday each week
For ten years, Resource Global has run on donor generosity. That works, but it is not sustainable as we move toward our goal of 25 cohorts in global cities around the world by 2030. My dream has been to build a for-profit company that funds Gospel work, and AI is the moment where this becomes possible for a small team with the right idea.
This past month has been one of the most consequential in AI history. Anthropic announced a security model so powerful they refused to release it publicly. Google made professional video creation free. Q1 venture funding hit $297B with 81% going to AI. The platforms are getting cheaper, the tools are getting more accessible, and the business models are getting clearer.
Three stories below, then one product idea worth wrestling with.
— Tommy
What's Launching
Building blocks. New platforms, models, and capabilities that could become the foundation for products.
Google I/O
Gemini Omni: professional video creation just dropped to zero cost
Gemini Omni Flash, announced Tuesday and called "Nano Banana for video" by Google itself, lets anyone turn text, image, audio, or video inputs into a polished video. Includes a personal Avatar feature: a digital version of yourself that looks and sounds like you, so you can generate videos starring you without ever filming.
Rolling out now to Google AI subscribers. Critically, also rolling out free on YouTube Shorts and the YouTube Create app this week.
Builder angle
Video creation just became a commodity. The product opportunity is not "AI video generator" (commoditized). It is in the layer above: who packages this for a specific audience, in a specific language, with a specific workflow? Think: AI sermon prep for under-resourced pastors. AI explainers for cohort members in Bahasa or Mandarin. Localized content engines for ministries that cannot afford production teams. The infrastructure is now free.
Cursor
Frontier coding at 1/10th the cost. The economics just changed.
Cursor's Composer 2.5, released this week, matches Anthropic's Claude Opus 4.7 and OpenAI's GPT-5.5 on coding benchmarks at less than $1 per task vs $11 for the frontier models. This continues a pattern: capability is no longer scarce, but cost-efficient capability is the new edge.
Builder angle
Margin math just shifted. A product built on frontier models last year would have been unprofitable serving a mission market (low-cost developing world). The same product on Composer 2.5 economics is viable. This opens up a category of "frontier-quality, mission-affordable" AI products that could serve markets in Africa, SE Asia, and Latin America at scale. Worth a conversation with the Jakarta team about which workflows we could build on this stack.
Anthropic · April 7
Claude Mythos: too dangerous to release. The cybersecurity floor just shifted.
This is the biggest story of the last month, and most people outside tech missed it. Anthropic announced Claude Mythos Preview in early April: a frontier model so capable at finding software vulnerabilities that the company decided not to release it. Mythos autonomously discovered thousands of zero-day vulnerabilities across every major operating system and web browser, including a 27-year-old bug in OpenBSD and a 17-year-old bug in FreeBSD. Engineers with no security training were able to use it to generate complete, working exploits.
Instead of public release, Anthropic launched Project Glasswing, a consortium with Microsoft, Google, Apple, AWS, Cisco, JPMorgan Chase, the Linux Foundation, and others, using Mythos to find and patch vulnerabilities before bad actors get similar models. Anthropic estimates that comparable capabilities will emerge from other labs within 12 to 18 months.
Builder angle
Two implications. First, any AI product we build needs serious security review from day one. The window between a vulnerability being found and weaponized is collapsing from weeks to hours. Second, this is a watershed moment for what AI is capable of. Anthropic chose restraint when they could have monetized. That posture of "we have power but we are choosing to steward it carefully" is exactly the kind of company a faith-driven venture should partner with. Project Glasswing also shows what AI for the common good can look like when frontier labs cooperate instead of compete.
Where the Money is Moving
Funding signals. What investors are paying for tells you what real customers will pay for, eventually.
Pattern
Vertical AI is beating horizontal AI. Niche workflows are the winners.
$18.8 billion has gone into AI startups founded since early 2025, but the pattern of winners is clear: vertical, workflow-specific tools that own one job in one industry. Rebar (HVAC quote generation) raised $14M Series A and doubled ARR in six weeks. Fazeshift (AI for accounts receivable) raised $22M in May. Performativ ($5.5M for wealth management) and Marloo ($10M for advisor workflows) are smaller but instructive.
The "AI for everything" generic products are losing. The "AI for this one painful workflow" products are getting funded and finding revenue fast.
Builder angle
The lesson for a Gospel-funding venture: do not build "AI for ministries." Pick one workflow. AI for small church accounting. AI for missionary support letter drafting. AI for translating sermons into 50 dialects. AI for grant writing. Each of these is small enough to dominate, big enough to fund. The vertical strategy is also more defensible against the giants.
Enterprise
Anthropic just won the audit firm market. 800,000+ professionals now use Claude.
In three weeks Anthropic has landed KPMG (276,000+ employees), PwC (hundreds of thousands), and SAP integration. This tells you where the real enterprise budget is moving: professional services, finance, and ops are buying Claude as the default reasoning layer.
Claude Code alone hit $2.5B in annualized revenue by February 2026. Anthropic is at ~$30-40B annualized revenue. OpenAI at ~$25B. The market is real and bigger than most people think.
Builder angle
Two relevant signals. First, building on Claude as infrastructure is now safer than ever because it has enterprise lock-in. Second, the consulting world (PwC, KPMG, Deloitte) is about to deploy AI tools to every client they serve. That means the SMB and mid-market space they neglect, especially in developing countries, is wide open for purpose-built tools that fit local context. That is our cohort members' actual market.
One Product Idea This Week
AI tools for cohort members to fund their own Gospel Action Plans
Every cohort member completes a Gospel Action Plan. Many of them in Indonesia, Malaysia, and Singapore use for-profit startups to fund Gospel work. What if Resource Global built a small suite of AI tools that helped cohort members run their businesses better, with a portion of revenue flowing back to RG?
Examples: an AI document and proposal tool for cohort members raising funds. An AI translation service for cross-cultural ministry content. An AI sermon and Bible study assistant in local languages. Each one solves a real pain in our network. Each one could become a small business with revenue share back to RG.
Hypothesis worth testing: A vertical AI product built for and with our cohort network would have built-in distribution (the cohort), built-in product feedback (real users with real businesses), and built-in mission alignment. That is a much better starting position than "another AI startup."
Business Models Working
Revenue patterns. What people are actually paying for, with real numbers behind it.
Price the Market
B2B AI tools are settling at
$20-25 per user per month (Claude for Small Business, ChatGPT Business). That is the ceiling we should plan for if we build for cohort members and small ministries. Higher than that and we lose them. Build the value math around saving them one hour per week.
Anthropic
Two Tier Model
Power users will pay
$100/month (Google AI Ultra, ChatGPT Pro, Claude Max). That's the tier for serious entrepreneurs, analysts, and creators in our network. Any product we build should likely have a free tier, a $20 tier, and a $100 tier. The middle is where most of the revenue lives.
CNBC
Charge for Outcomes
PwC: insurance underwriting that took 10 weeks now takes 10 days using Claude.
That is the kind of outcome customers will pay 10x more for. Per-seat pricing is dying. The new model is "we will deliver this outcome, here's what it costs." Think about pricing our future product per Gospel Action Plan, per translation, per sermon, not per user.
PwC
Capital Available
Vertical AI Series A rounds are averaging
$51.9M, 30% higher than non-AI. Seed rounds for AI startups get a 42% valuation premium. If we want institutional capital eventually, the math has never been more favorable. But: investors want a clear workflow, not "AI for ministry" framing.
Qubit Capital
Watch & Apply
Anthropic's Chicago workshop on
May 14 is free, with a one-month Claude Max included. Donna should look into whether any of our Chicago cohort members or RG donors can attend the next round. The full city list: Tulsa, Dallas, Hamilton Township, Baton Rouge, Birmingham, Salt Lake City, Baltimore, San Jose, Indianapolis.
Anthropic
Government Pays
OpenAI's Malta deal is a template. Every Maltese citizen who completes a national AI literacy course gets free ChatGPT Plus. Indonesia, Singapore, and Malaysia will likely follow within 12 months. If we build AI training content with Gospel undertones, there is a real path to government-funded distribution in countries where our cohorts already operate. Worth a conversation with Melissa Tan in KL and the Jakarta team.
OpenAI
Pick a Side
OpenAI is putting ads in free ChatGPT. Anthropic has pledged ad-free permanently.
For a Gospel-aligned product, we build on Claude. Same values, different posture. This is not just a technical choice; it's a brand alignment that will matter to donors and cohort members who care about how their data is used.
Anthropic news
10x Cost Drop
Cursor's Composer 2.5: $1 per task vs $11 on frontier models. Translation: a product that would have cost $300/month to operate last year now costs $30. That is the entire business model unlock for serving cohort members and ministries in developing markets. This is the most important number on this page if we want to serve Jakarta, Nairobi, Cairo, São Paulo at scale.
Cursor
Free This Week
Gemini Omni is free on YouTube Shorts and YouTube Create app starting this week. Cohort members can produce professional video for their Gospel Action Plans at zero cost. Worth a Slack to all city directors: try this with your cohort this week. It's a 5-minute experiment that could change what they think is possible.
Google
What I'm Reading
Sources I rely on, with builder lens.
Daily News
The Rundown AI. 5 min/day, 2M readers, free. Best general AI source. therundown.ai
Funding
Crunchbase News and The Information for funding rounds. Stratechery by Ben Thompson for business model analysis.
Builder
Lenny's Newsletter for product and growth. Latent Space by swyx for technical builder perspective.
Sources
Direct: anthropic.com/news, blog.google, openai.com/news. Skip the commentary, read the source.
That's issue two. Talk to you next Friday.
Tommy