Last chance to lock in 2025 rates

At Future Works, our rates are increasing by 20% for 2026. As a thank you for being part of the Lighthouse community, I'm giving you the chance to lock in our current pricing for all of next year.

We now ship critical initiatives 2x faster than the market standard. This means we can execute your 2026 roadmap in half the time.

If you book your roadmap session before December 31, you will lock in our current 2025 pricing for all of next year.

If you have been waiting to build, this is the time.

In today's lineup: 

Robots

• Your new moat is a private AI factory

• Real ROI from Gen AI

• NotebookLM or Perplexity for your research workflow?

People

• Tim Latimer engineers the power for the AI age

Love

• The antidote to slow execution

• Transformative Reads: Measure What Matters

Top Insights

• Enterprise AI is moving from open-ended chat to reliable, protocol-driven task execution.

• Engineering agents can now autonomously refactor entire legacy systems.

• Record capital is flooding the physical backbone of the intelligent web.

• China is bypassing the blockade with domestic machinery for advanced silicon.

• Washington is finally treating compute like a utility, slashing permit times for data centers.

• The carbon cost of intelligence now matches the entire footprint of New York City.

Enterprise AI is the fastest-scaling software ever

Enterprise AI spend jumped from $1.7 billion to $37 billion in 18 months. Menlo Ventures reports it now captures 6% of the global SaaS market. This capital follows execution. OpenAI data shows 800 million weekly users have moved past testing. Frontier firms send 7x more messages to Custom GPTs than average companies.

Execution priorities

• Deploy agents for multi-step tasks. Leaders process 20% of messages through custom workflows rather than basic chat.

• Stop building custom models when vertical SaaS works. The $37B surge confirms the market demands specific outcomes over raw tools.

• Enforce a standard toolchain. Security protocols must travel with the data as you scale from pilot to production.

• Treat AI consumption as COGS. Monitor token usage per workflow to ensure automation costs less than the human labor it replaces.

Real ROI from Gen AI

How human-centric AI delivers business results

• Amazon saved 4,500 developer-years by upgrading 30,000 apps with Amazon Q, generating $260M in efficiency gains.

• Wayfair cut physical studio costs by deploying Google Vertex AI to generate "virtual photography" for its massive asset catalog.

• Pfizer saved 16,000 research hours annually by using AWS Bedrock and Claude 3 to automate scientific review and patent drafting.

Tool Spotlight

Your weekly briefing on tools that create competitive leverage

GPT Image 1.5 or Nano Banana Pro?

GPT Image 1.5 

Released Tuesday as OpenAI’s "code red" response to Google, this update brings precise editing tools to all users.

Best for: Ops and Design teams needing rapid, iterative edits without technical friction.

Choose if:

• Specific errors like text or lighting must be fixed without regenerating the entire asset.

• A dedicated studio interface is required to accelerate design cycles by 4x.

• Legible text rendering is needed directly inside generated graphics.

Nano Banana Pro 

Google’s model recently overtook OpenAI in market share by solving the "consistency problem" in generative media.

Best for: Innovation leaders requiring strict visual consistency across brand assets.

Choose if:

• Granular control over facial likeness and color tones is necessary across multiple generations.

• The priority is using the model currently topping LMArena benchmarks for visual fidelity.

• Complex compositions require precise post-production adjustments that maintain layout integrity.

3 alternatives to explore

• Adobe Firefly: The safest choice for enterprise commercial use with IP indemnification.

• Midjourney v6.1: Still the standard for high-end artistic composition and texture.

• Flux.1: An open-weight model for teams that need self-hosted privacy.

Together with Wonder Tools

The independent guide to smarter tools

Join 80,000+ professionals who rely on Wonder Tools for a free, 5-minute weekly briefing on the most useful apps and AI tools. Written by a former Time Magazine journalist, it cuts through the hype to save you and your team hours of trial and error.

Each concise issue delivers:

• Clear, honest reviews of practical productivity and creativity tools.

• Actionable ways to use them for your own work.

• Get an independent take on what's actually worth your time.

Subscribe for free or upgrade for special perks. This includes a bundle of free pro AI & other premium tool subscriptions.

Transformation Champion

Bob Mumgaard is moving fusion from the lab to the construction site

Mumgaard stopped treating fusion like a physics problem and started treating it like a construction project. While the industry fixated on "net energy" metrics, he focused on permits and supply chains.

The results are concrete: his Boston test site is 75% complete, and he has broken ground on a grid-scale plant in Virginia backed by a Google power purchase agreement. Mumgaard successfully bypassed the slow-moving Nuclear Regulatory Commission for faster state approvals and partnered with Google DeepMind to solve engineering bottlenecks. Now, he is pushing for a "CHIPS Act for fusion," warning Congress that "spending billions now avoids spending ten times that to recover later."

Regain velocity through federation

Corporate centralization destroys speed. Leaders often obsess over a single "source of truth," but this creates a bottleneck where the slowest approver dictates the pace. You end up with a unified mess instead of a nimble organization.

Three shifts to regain velocity:

1. Federate your AI. Let business units build their own models on a shared data layer.

2. Standardize protocols, not tools. Enforce how systems connect via APIs, but let teams pick the best software for their work.

3. Capture knowledge collisions. Structure meetings to find friction points rather than forcing rigid consensus.

Get the full federated operating model in my book, 100x.

Transformative Reads

One book, handpicked from my conversations with friends, industry leaders, and tech innovators:

Co-Intelligence by Ethan Mollick argues that we must stop treating AI like software and start treating it like a talented, uneven intern. He reveals the "jagged frontier" where models ace complex reasoning but fail simple tasks. Don't automate blindly. Use AI to challenge assumptions, act as a sounding board, and scale human expertise.

Perfect for: Operations and innovation leaders.

In Culture

As OpenAI chases a reported $750B valuation to 'advance human knowledge,' the line between noble explorer and assimilating hive mind is getting a little blurry...

THANK YOU!

Lighthouse takes us many hours to create, and we don’t run ads or charge anything. We are dedicated to making it the best AI transformation newsletter possible.

We just ask for one thing: tell one friend about Lighthouse today!