AI-native engineering · Fixed price
Your team has Claude Code. Getting more finished work out of it takes clear tasks, the right codebase context, and tests and review that verify the result. We build those practices into your team’s daily work in six weeks, then measure the change.
Two-week assessment, $25,000. Six-week embed, $35,000. A one-year Fulcrum license included with either. One signature, no procurement cycle.
Fixed scope, fixed price, one signature. Both start within two weeks of it.
Book a call about the Assessment
One team, one real backlog. We measure how your engineers and their agents actually work today, audit the repos for what agents need and do not find, and demonstrate the difference on one of your own tickets.
Charles’s time: up to 40 hours over the two weeks, about 8 of them live with your team.
Fulcrum included. A one-year license to install and use Fulcrum in your environment. After that, $1,199 per year, only if you choose to keep it.
One team of up to six engineers. Week one is the baseline. Weeks two through five build the guidance, task workflows, tests, and review practices into your repos, pairing on real tickets. Week six measures it and hands it over.
Charles’s time: up to 110 hours over the six weeks, about 38 of them live with your team.
The guarantee. The second half of the fee is invoiced only if the measured lift at week six is at least 3× your week-one baseline.
Fulcrum included. A one-year license to install and use Fulcrum in your environment. After that, $1,199 per year, only if you choose to keep it.
Everything on this page in a document you can hand to the person who signs: scope, weeks, deliverables, price, and terms. It is tagged for screen readers.
vantalect-engineering-leverage.pdf
Two fixed-price ways to install AI-native engineering in one team: a two-week Assessment at $25,000 and a six-week Embed at $35,000, half of it tied to a measured 3× lift.
Download PDF of Engineering LeverageValue the extra work your team can actually deliver. Then compare it with the full $35,000 Embed fee.
Engineering Leverage
Even a modest gain in usable capacity can be worth more than the engagement. You keep the team—and give it more room to deliver.
6 engineers at $180,000 annual loaded cost each. A 20% usable capacity gain, less $1,000 per month in added tools.
Loaded cost includes compensation, benefits, and overhead. Enter dollars with or without commas. Values must be nonnegative; model up to six engineers. Count capacity you can actually use after review and coordination.
This is the cost-equivalent value of additional output, not cash recovered. The example is not a forecast. The engagement’s 3× condition measures execution-factor lift; it does not establish a 3× gain in team capacity.
Monthly value = engineers × annual loaded cost ÷ 12 × usable capacity gain − added monthly tools. Fee-equivalent time = $35,000 ÷ positive monthly value.
| Month | Capacity value | Fee |
|---|---|---|
| 0 | $0 | $35,000 |
| 3 | $51,000 | $35,000 |
| 6 | $102,000 | $35,000 |
| 9 | $153,000 | $35,000 |
| 12 | $204,000 | $35,000 |
Six weeks to install and measure. Then the return accumulates. The charts start after the week-six measurement and assume the monthly benefit holds. They compare cumulative benefit with the full $35,000 fee, before taxes and financing. Your team’s time during the engagement is additional.
See the scope and result conditionsThe work behind the assessment
This is the work of building AccelaStudy® AI: one founder directing AI agents across the product and the tools behind it. A 98-second look at the work, the measurement, and the scale.
Charles has directed agents through a large, connected body of engineering work and recorded the tasks. He brings that practical experience to the questions your team needs answered: what context is missing, what checks are reliable, and what improvement the records support.
Repository, code, test, and content counts show scale. They do not establish software quality or financial return on their own. Senior-engineer hours are estimates; the film’s COCOMO figure is a model-based comparison. Your team’s result gets measured separately.
Inspect the published measurement recordFebruary 22–September 22, 2026: seven months of Charles’s work on AccelaStudy AI. The film uses energetic instrumental music and sound effects throughout, with rapidly changing text and visualizations. It has no spoken narration.
Time and orchestration: more than 100 hours a week; approximately 3,400 hours worked personally. Charles directs dozens of agents in parallel through orchestrating agents. The project is self-funded, without outside investment.
AI usage: estimates from development logs total 800 billion tokens: 500 billion Claude tokens and 300 billion GPT tokens. Monthly estimates are 85 billion Claude tokens and 72 billion GPT tokens, with approximately 245,000 and 300,000 responses respectively. Educational content generation is excluded from those development estimates.
Software and documentation: 188 repositories; 4,751,501 lines of code across Python, TypeScript, Swift, and Kotlin; 119,556 automated tests; 12,852 Playwright end-to-end tests; 32,506 pages and 8,939,059 words of documentation. Animated tiles, code, and dots make the scale visible.
Learning library: 1,534,347 practice questions, each represented by a point of light; 26,385 lessons; 378,356 flashcards; 3,248 case studies; 8,032 symbolic problems; 3,652 worked examples; 3,838 glossary terms; 2,079 simulations; 2,196 labs; 97 interactive tools; and 107 learning activities. An AI tutor connects to the learning system.
Patent portfolio: the film describes AVIAN as patent pending, with 49 application documents including 23 being filed at the time, 302 inventions, 1,158 claims, and 407 figures. The snapshot does not establish that all 49 were filed or granted.
Measurement: 3,745 logged tasks represent 102,955 estimated senior-engineer hours, compared with approximately 3,400 hours of Charles’s own work. At 2,080 working hours per year, the estimated task work corresponds to 49.5 years. The film also shows a COCOMO estimate of 1,451 person-years based on software size; this is a theoretical comparison, not measured elapsed labor or a client result.
Closing: one person can direct a body of work that once required an organization. The film includes a forward-looking aspiration of more than 100 years of estimated work by the one-year mark; it is not a result verified in this snapshot. It ends on “The Scale of the Work” and “Seven months.”
Aggregate execution factor: senior-engineer-equivalent hours per task, divided by agent session minutes, summed over 3,552 tasks logged since February 22, 2026 (median task 22.5×). Every row is published and recomputable at charlessieg.com/leverage. Measured on our own build, not promised for yours. Yours gets measured.
Because the work has a shape.
The same weeks, every time, which is why it is fixed price.
Want to walk through these weeks for your own team?
Book a 15-minute call about these weeksEvery deliverable is an artifact in your repositories or your inbox.
None of them is a promise in a deck.
Fixed price does not mean unbounded.
The cap is in the statement of work, and so is what we need from you.
Over two weeks. About 8 of them are live with your team; the rest are the audit, the plan, and the agents doing the build work under Charles’s supervision.
Over six weeks, roughly 18 hours a week. About 38 of them are live with your team: the weekly session, the pairing, the office hours, and the readouts.
By written change order, agreed before the hour is spent. Remote by default; an on-site day is priced separately.
Every meeting is on the calendar before week one.
Nothing else is asked of your engineers except the pairing.
Sponsor and engineering leads. Goals, access, the measurement contract, and what happens in the two weeks.
The one class: estimating senior-engineer hours, what counts as a task, and the MCP server and CLI that capture it.
Drop-in while the team baselines.
One ticket from your backlog, done AI-native in front of everyone, questions throughout.
The baseline number, the audit findings, the installation plan, and what six weeks would change.
Week one is the Assessment’s week one: the kickoff, the logging class, and office hours. Then:
The number so far, what went in this week, what is blocked.
A class on the week’s theme with a live demonstration: constraints, gates, workflows, tooling.
With the engineer who owns the repo being worked. This is where each guidance file is proven on a real ticket, and where the teaching actually happens.
Drop-in, all six weeks.
The measurement review, a 2-hour handover workshop, and the closing sponsor readout.
A pavilion is a structure built for one occasion and handed over when it ends. Every engagement gets one: a private online workspace, set up in week one for your sponsor, your engineers, and us, holding everything the engagement produces and everything it needs from you.
The calendar of sessions. Every recording, with its transcript and a reviewed summary for the people who were not in the room. The audit, the plan, the readouts, and the playbook, each a version your sponsor accepts. The discussion around them. The decisions, with their reasons. The action items and the obligations on both sides. The measurement panel, wired to your Fulcrum. At the end, one archive of all of it, which is yours. Read about Pavilion.
To see a workspace carry a whole engagement, read the two StrataForge sample case studies: the security hardening and the cost optimization, each told week by week from what its workspace recorded.
Pavilion is a product of Renkara Media Group, Inc., operated for the engagement and exported to you at its close.
Nothing exotic, and nothing you have to buy.
Everything the engagement produces is recorded, transcribed, and left with you.
Installed in your environment under the included license. Every task the team logs, the baseline, and the week-six number live here, and the record stays yours.
Guidance files, gates, workflows, and the playbook are committed to your repositories and run in your continuous-integration system. Nothing is hosted by us.
On your subscription, in your accounts. The live ticket, the pairing, and every workflow are run with the tool your engineers already have.
Every session is held on Zoom and recorded with your permission. Zoom’s transcripts and recaps are delivered with each recording, so the people who were not in the room get the whole meeting, not a summary of it.
Set up for your engagement in week one. The calendar, every recording with its transcript and summary, the documents your sponsor accepts, the discussion, the decisions, the action items and obligations, and the measurement panel, in one place you keep afterward.
Scheduling, approvals, change orders, and invoices go by email, so the commercial record of the engagement is in your inbox, not in a portal you lose access to.
Fulcrum is the instrument behind the 31× number: a leverage record kept at the time of the work, not reconstructed afterward. Every task carries the senior-hours it would have taken and the agent minutes it did take, so the week-six number is measured with the same instrument as week one.
Both engagements install Fulcrum in your environment under a one-year license. After the first year, keep it for $1,199 per year, or do not. The record you built stays yours either way. Read about Fulcrum.
Between February and August 2026, one engineer and Claude Code built and launched an adaptive learning platform. Everything below is measured, not estimated, and most of it is public.
The engineering practices behind those numbers are what we put in place: 57 nested guidance files that constrain every agent session, pre-commit gates that refuse untested code, a fleet of internal tools the agents operate, and a leverage record kept at the time of the work, not reconstructed afterward.
Read the thinking: The Leverage Factor, Defending the Numbers, and Agentic Coding and Decision Fatigue, republished on the blog from charlessieg.com.
Charles Sieg. Not a team that learned this from a course; the person who did the work above. Building software since 1997. Shipped AccelaStudy on the App Store’s opening day in 2008 and grew it to 30 million downloads across 300 published apps. Cloud architect and platform engineer for Fortune 1000 clients across healthcare, pharma, and financial services. Sole inventor on every patent in the portfolio. Both engagements are delivered by Charles personally, inside your repos, with your engineers.
No. The work happens in your repositories, your accounts, and your model subscriptions. Nothing is copied out, and nothing of ours is left behind that you do not own.
You do. The guidance files, gates, workflows, and playbook are yours outright. We keep nothing but the measurement, with your name off it unless you say otherwise.
For every task, we estimate the hours a senior engineer familiar with the codebase would need, and we record the agent session minutes it actually took. Summed over 3,552 tasks, the ratio is 31. The median task is 22.5×. It is an upper-tail record of one operator with full product authority and no handoffs, which is why your number gets measured rather than assumed.
Because the work has the same shape every time and because hourly engagements go through procurement. A fixed-price sprint with a defined deliverable is a purchase a VP can make this week.
Then you pay half. The second half of the embed fee is invoiced only if week six measures at least 3× week one. You keep everything that was installed either way.
Up to 40 hours in the Assessment and up to 110 in the Embed, with about 8 and 38 of those hours live with your team. The caps are written into the statement of work. Anything beyond them is a written change order at $450 per hour, agreed before the hour is spent.
Both, and pairing does the heavier lifting. The Assessment has one class, forty-five minutes on how to log leverage. The Embed adds a ninety-minute session each week on that week’s theme, with a live demonstration, and two or three pairing sessions a week with the engineer who owns the repo. Every session is recorded for the people who could not be in the room, if you allow recording.
No. The engagement is about how your engineers and their agents work. Cloud platform, language, and repository host do not change it, though AWS is where we have the deepest history.
Not during the engagement. Both engagements include a one-year license to install and use Fulcrum in your environment. After that it is $1,199 per year, only if you choose to keep using it. There is nothing to cancel, and the leverage record you built stays yours.
Two weeks and $25,000 for the diagnostic, six weeks and $35,000 for the result.
The AWS bill grew faster than the business, and nobody can say who owns which line.
Read the engagement
An audit is coming, a customer sent a security questionnaire, or nobody can say for certain who can reach production.
Read the engagement
A critical production workload has grown under deadline pressure, and nobody has reviewed its reliability, security, performance, and cost together.
Read the engagementTell us the team, the stack, and what your board asked. You will hear back from Charles within one business day, and the first call is fifteen minutes.
Claude and Claude Code are trademarks of Anthropic, PBC. GitHub Copilot is a trademark of GitHub, Inc. Cursor is a trademark of Anysphere, Inc. Zoom is a trademark of Zoom Video Communications, Inc. Vantalect is not affiliated with or endorsed by any of them.