🎮 The Next Input — Issue #149

Sponsored by

⚡ The Briefing — 60 sec

• Cursor admits its new coding model was built on top of Moonshot AI’s Kimi Paying premium money for dressed-up open models is going to become a very real conversation. Cursor said Composer 2 started from an open-source Kimi base and only later acknowledged that publicly, which is exactly why buyers are starting to look harder at what they are actually paying for.

• Are AI tokens the new signing bonus or just a cost of doing business? “Competitive comp” now apparently includes a mountain of inference credits. If a role needs a quarter-million in tokens just to function, that is not a quirky perk anymore — that is a cost structure.

• AI-powered cows show there is no end to artificial intelligence’s reach The cowgorithm is strong with this one. AI collars are already being used to move livestock, monitor herd health, and cut labour on farms, which is a good reminder that AI is not stopping at chatbots or coding tools.

🛠️ The Playbook — The AI Spend Accountability Engine

Mission
Track where AI is genuinely creating leverage versus where you are just paying premium prices for repackaged models, rising token burn, and vague promises.

Difficulty
Intermediate

Build time
3–5 hours

ROI
Lower wasted spend, clearer vendor decisions, and faster identification of workflows where AI is actually worth the money.

0) Why This Matters

The AI market is starting to mature in an awkward way.

On one side, buyers are discovering that some “frontier” products may be built on top of open or commercially available base models with extra training and packaging layered on top. Cursor’s VP of developer education said Composer 2 “started from an open-source base,” after users identified signs it was based on Moonshot AI’s Kimi 2.5, and the company later conceded it should have disclosed that up front.

On the other side, usage itself is becoming a budget line. TechCrunch reports that AI tokens are increasingly being discussed like a compensation lever, which means model consumption is no longer just infrastructure — it is becoming part of how teams recruit, equip, and justify work.

And then there is the third signal: AI keeps spreading into sectors people barely associated with software a few years ago. Halter’s smart cattle collars use sound and vibration cues to move cows, while also monitoring animal health and fertility, with expansion underway in Australia and the U.S.

So the operator move is simple:

• know what model you are actually buying

• know what the workflow really costs

• know whether the output is worth the burn

• know where AI is creating real operational edge

1) Architecture

2) Workflow

1. List every workflow currently using AI, including chat tools, coding tools, automations, and connected apps.

2. Record what model or vendor is underneath each workflow and whether it depends on another base model.

3. Track the real cost of usage, including seats, tokens, API calls, and any hidden support overhead.

4. Review outputs for usefulness, accuracy, and the amount of human correction still required.

5. Compare total spend against time saved, quality improved, or revenue unlocked.

6. Cut the workflows that are mostly wrapper tax and expand the ones producing measurable leverage.

3) Example Prompts

Vendor Transparency Prompt

You are reviewing an AI product for procurement risk.

For the product below, identify:
- the likely underlying model or dependency stack
- whether the vendor appears to be a wrapper, fine-tuner, or original model builder
- what questions should be asked before purchase
- the top 5 transparency risks

Product:
[insert product]

Workflow Costing Prompt

You are calculating the real cost of an AI-assisted workflow.

For the workflow below, estimate:
- tool subscription cost
- token or API cost
- human review time
- rework or correction time
- total weekly cost

Then explain whether the workflow appears:
- high ROI
- uncertain ROI
- negative ROI

AI Usefulness Audit Prompt

You are auditing whether an AI workflow is genuinely useful.

Review the workflow and score:
- speed gain
- output quality
- reliability
- human edit load
- business value

Return:
1. overall score out of 10
2. biggest strength
3. biggest weakness
4. recommendation: expand, fix, or cut

Agriculture / Field Workflow Prompt

You are designing an AI workflow for a non-desk industry.

Given the process below:
- identify where sensing or monitoring data is created
- identify where AI can improve decisions
- identify where humans must stay in control
- identify the top 5 operational risks

Process:
[insert process]

4) Guardrails

• Never assume a premium AI product is built from scratch.

• Separate model quality from brand positioning.

• Track token burn at workflow level, not just account level.

• Count human correction time as part of total cost.

• Review whether AI is improving throughput or just shifting effort around.

• Do not expand a workflow until you can prove it beats the old method.

5) Pilot Rollout — 3 hours

1. Pick three live workflows currently using paid AI tools.

2. Map the vendor, model dependency, and total weekly cost for each one.

3. Measure the actual output quality and human rework required.

4. Build a simple dashboard comparing spend, time saved, and business value.

5. Classify each workflow as expand, fix, or cut.

6. Use the results to create a standing review process for all future AI tooling.

6) Metrics

• Cost per workflow run

• Tokens consumed per employee or team

• Human edit time per AI output

• Percentage of workflows with positive ROI

• Vendor transparency score

• Number of duplicate or overlapping AI tools

• Weekly spend avoided through workflow cuts

Pro Tip: The fastest way to waste money in AI is to measure the model and ignore the workflow wrapped around it.

🎯 The Arsenal — Tools & Platforms

• Airtable · track workflows, owners, and expansion or cut decisions · Airtable

• Google Sheets · lightweight ROI model for seat cost, token burn, and edit time · Google Sheets

• Cursor · useful coding workflow tool, but also a reminder to inspect what is underneath the hood · TechCrunch coverage

• API billing dashboards · ground token-spend discussions in actual usage instead of vibes · TechCrunch coverage

• Field AI systems · proof that AI ROI can show up far beyond desk work, including livestock management and health monitoring · News.com.au coverage

Copy-paste prompt block:

You are helping me build an AI Spend Accountability Engine.

For the workflow below:
1. identify the AI tools and vendors involved
2. identify the likely model dependency stack
3. estimate subscription, token, and review costs
4. estimate weekly hours saved
5. assess output quality and human edit load
6. classify the workflow as expand, fix, or cut
7. list the top 5 risks of overspending on this workflow

Workflow:
[insert workflow here]

Return the answer in markdown with sections for:
- Workflow summary
- Tool stack
- Cost structure
- ROI assessment
- Human review load
- Risks
- Recommendation
- Metrics

💡 Free Office Hours

If you are trying to work out whether your AI stack is creating real leverage or just expensive noise, I run free office hours to help map the workflows, the spend, and the fastest path to a cleaner setup.

Book here: https://calendly.com

Experts Would Invest $100,000 in This Alternative Now

A new Knight Frank report made an unexpected declaration. It revealed that 44% of family offices are investing more in residential real estate now. And, you don’t need to be Warren Buffet to see why.

Since 2000, residential real estate outperformed the S&P 500 by 70% in total returns. It’s the only asset that pays you to own it, grows while you sleep, and shields your gains from the IRS. 

That’s why you need mogul. It’s a real estate platform that lets you invest in institutional-grade rental properties. You get monthly rental income, capital appreciation and tax benefits without a down payment or 3 a.m. tenant calls. In fact, over 20,000 investors have joined. 

Here’s Why:

• Tax Benefits

• +7% annual yields

• 18.8% avg annual IRR

TLDR: You can invest in high quality real estate for a fraction of the cost. Why wait?

Unlock This Alternative Now

Past performance isn't predictive; illustrative only. Investing risks principal; no securities offer. See important Disclaimers

🕹️ Game Over

AI is reaching everywhere now. Good. That just means the adults need to start checking the bill.

— Aaron Automating the boring. Amplifying the brilliant.

Subscribe: link