[No.121] AI Infrastructure-Oriented Shift Practical Pack — 1-Page Inference Cost & Proposal Attachment Template
"Senior engineers are in short supply" and "hiring seniors is slow" can both be true at the same time.
In 2026 market analysis, generalist senior SWEs are in oversupply, taking over 60 days to fill, while AI infrastructure specialists are rare and filled in a few weeks, a polarization organized by Kore1 and others.
Even with the same job title, what you are selling has made the market a different beast.
This article is a deep dive that doesn't just end the polarization as news, but summarizes it as a 1-page inference cost and proposal template to take one step from generalist to infrastructure-oriented.
What this article includes
How to read polarization — Generalist 60+ days vs. Infrastructure few weeks, the context of 29% trust
1-page inference cost template — Aggregation format to attach to internal/client proposals
3-line improvement proposal template — How to write about caching, small models, and batching
Cover letter for proposal attachment — Short text for bosses/clients
How to fix your LinkedIn job section — A pattern for adding one word to shift from generalist to infrastructure-oriented
Positioning self-assessment — Visualizing "what you are selling" with 10 items
30-day roadmap — From 1-page cost to proposal attachment
What you can take away from this article
AI only compresses typing, and inference, GPU, and monitoring that crush how things break first command a premium. The Stack Overflow 2025 survey cites 29% trust in AI output and increased debugging load, which can be read as a context where the value of arbitration and infrastructure is rising. The minimum action for this week is to summarize internal inference costs on one page and attach it to a proposal.
One move to try this week: Fill out the "1-page inference cost template" below for the last month and attach it to a proposal for your boss or client.
The contours of polarization — How to read the numbers
Using Kore1's analysis as a clue, the following contrast is shown (this does not guarantee individual career change results).
Senior Generalist SWE
Oversupply, 60+ days to fill, mentions of flat to declining compensation, demand compressed by AI coding tools.
AI Infrastructure Specialist
Rare, filled in a few weeks, mentions of ranges like $245K–$450K+, demand increased by tool proliferation.
Collapse of trust
Stack Overflow 2025 survey cites 29% trust in AI output and increased debugging load.
Even for the same "senior" role, one can read it as the market for competent typing being a different beast from the market for mechanisms that don't break in production + cost.
How to read polarization — 3 branches (framework)
1. What is being reduced?
General-purpose roles focus on 'time spent writing code,' while infrastructure roles focus on 'investigation, costs, and latency when things break.'
2. What can be expressed in numbers?
General-purpose roles focus on feature delivery, while infrastructure roles focus on inference costs, SLAs, and error rates.
3. Who feels reassured?
A scenario where clients and management are starting to buy 'documents that show costs and monitoring' rather than 'working demos.'
Common Misconceptions and Rebuttals
Misconception 1: 'Moving to infrastructure is easier'
Rebuttal: It is not about changing your job title, but whether you can discuss costs and reliability using numbers that marks the turning point.
Misconception 2: 'General-purpose seniors are no longer needed'
Rebuttal: Oversupply is a matter of days to fill a position. The analysis suggests that by changing one word in your position, the same experience can reach a different market.
Misconception 3: 'Knowing GPUs is enough'
Rebuttal: The point of secondary reporting is the set of inference costs, monitoring, and permissions.
Inference Cost One-Pager Template (For Copying)
Do not use tables; summarize on one page using label blocks. This can be used in internal dashboards or spreadsheets.
Reporting Period: (e.g., May 1, 2026 – May 31, 2026)
Total (Estimated): ( ) JPY / ( ) USD
By Model
(e.g., GPT-4.x): ( ) — Primary Use: ( )
(e.g., Claude 3.x): ( ) — Primary Use: ( )
(e.g., In-house/Open Weights): ( ) — Primary Use: ( )
By Function
(e.g., Customer Support Summary): ( )/month
(e.g., Code Review Assistance): ( )/month
(e.g., Batch Analysis): ( )/month
Insights (1 line): The highest cost is ( ), accounting for approximately ( )% of the total.
Data Source: (e.g., internal billing dashboard, manual aggregation)
Note (1 line): This is an estimate, not a finalized billing amount.
3-Line Improvement Proposal Template
Paste directly below the one-page cost summary.
Proposal 1 (1 line): (e.g., Implement caching for the summary API to reduce re-calls for identical queries)
Proposal 2 (1 line): (e.g., Route sub-tasks to smaller models)
Proposal 3 (1 line): (e.g., Level out peak tokens by batching at night)
Expected Impact (Qualitative, 1 line): (e.g., Estimated reduction of X-Y% in monthly costs, measurement required)
Next Experiment (1 line): (e.g., Compare token counts via logs over a 2-week A/B test)
Proposal Attachment Cover Letter (for managers/clients)
Subject: Inference Cost Summary and Improvement Proposals (1-page attachment)
Body:
Hello, this is (Name).
I have summarized the inference/API costs for the last month in one page (attached). The biggest cost driver is ( ).
I have included 3 lines of improvement proposals. I plan to test (Proposal 1) for two weeks first. If you approve, I would like to add this to the agenda for the next sprint.
How to update your LinkedIn job title (Add one word)
You don't need to change your job title immediately. Add it to the subtitle or the first line of your About section.
Current Job Title: (e.g., Senior Software Engineer)
Word to add (choose one): AI Infrastructure / Inference Cost / MLOps / LLM Operations / Production Inference
1-line profile example: Senior SWE — (Added word) and inference cost optimization. Recently (1-line summary of improvement proposal).
Items to feature: 1-page inference cost (confidentiality masked version) or a link to the improvement experiment README
Positioning Self-Assessment (10 items)
0-2 points per item. A total of 14 points or more is a benchmark for being able to speak from an infrastructure perspective.
Can state the inference cost numerically for the last month
Broken down by model
Broken down by function
Wrote 3 lines of improvement proposals
Have 1 line of test planning for improvements
Attached 1 page to proposals/for the boss
Have one infrastructure-related term on LinkedIn
Can touch on monitoring/error rates
Can explain permissions/production boundaries
Can explain the 29% reliability of secondary reports as 'that is why verification is needed'
30-Day Roadmap
Week 1 — Aggregation
Collect one month's worth of data via billing dashboard or manual entry
Fill out the 1-page inference cost template
Definition of done: Can state the total and the largest driver in one line
Week 2 — Improvement Proposal
Add a 3-line improvement proposal template
Decide on the next experiment in one line
Definition of Done: Can explain it orally to a manager in 30 seconds
Week 3 — Attachment
Send one page in the proposal attachment cover letter
Definition of Done: Receive one reply or agenda-setting response
Week 4 — Presentation
Add one word to LinkedIn
Add a 1-page mask version or README to Featured
Definition of Done: Can include the profile URL in job change or contract proposals
How to use by persona
Generalist Senior — Without changing your job title, add one word oriented toward infrastructure to "what you sell" using one page of cost. While the days to fill a position depend on the market, how you present the same experience changes.
Contracting — One page of cost + 3-line improvement proposal attached to client proposals. You can position yourself on the side of "reducing operational costs" during rate negotiations.
Side Hustle — Put the Before/After of a small inference improvement (one cache) in your README. It becomes one item in your professional portfolio.
Relationship with short videos
Short 0121 uses the hook of [Polarization] to convey the excess of generalists and the shortage of AI infrastructure in one go. The video is the map, and this article is for saving the 1-page cost, improvement proposal, cover letter, and LinkedIn content. Please start from the aggregation in Week 1.
Reference
Secondary information (Summary of Kore1/Stack Overflow 2025 Developer Survey around 2026-06). Market analysis stating that senior generalist SWEs are in oversupply with over 60 days to fill, while AI infrastructure specialists are rare with salaries of $245K–$450K+. Mentions research on 29% AI output reliability and increased debugging load. This does not guarantee individual job changes or salary.
Disclaimer
This article is a summary based on secondary analysis and a general cost organization template. Salary ranges and lead times vary significantly depending on research, region, and timing. Please do not disclose confidential internal figures as they are. Please take responsibility for your own career changes and unit price decisions.
Edo Tech Kawaraban — Survival Strategy in the AI Era