Aithon Blog
Aithon Engineering
Agents do the toil. Humans hold the gates. Real numbers every post.
How a Tiny Team Ships Like a Big One
We gave Claude hands: MCP from internal tool to product feature
How two small read-only MCP servers changed our engineering day-to-day - and why the same protocol ended up as a customer-facing product surface with OAuth discovery.
5 min readmcpclaudeawslambdaproductHow a Tiny Team Ships Like a Big One · The Product Track
44 use cases we never wrote
A sales team can't pitch what it can't enumerate. We mined public case studies - a customer's own and their competitors' - into a structured use-case library, and ranked 261 accounts against it.
4 min readai-engineeringproductgtmllmHow a Tiny Team Ships Like a Big One
Humans own the spec: how a retro complaint becomes shipped code
Every complaint, customer question, and retro item lands in one intake pipeline that sorts work into mechanical, design, and human lanes - and design work can't start until a person approves the spec.
7 min readagentsprocessclaudespec-drivenHow a Tiny Team Ships Like a Big One · The Partner Track
When to say no to an AWS program
Checking eligibility for an AWS partner program takes minutes; satisfying one can take a sprint. The three questions we price a yes with, and the one exception worth making.
4 min readawspartnerfundingstrategyHow a Tiny Team Ships Like a Big One · The Product Track
We built a context graph, then made ourselves justify it
Context graph, MCP, memory, agentic: the phrase on the slide keeps changing. The useful thing to publish is the internal argument that outlasts it, and where we decided the hype ends.
6 min readknowledge-grapharchitectureai-engineeringproductHow a Tiny Team Ships Like a Big One
We don't have a platform team. We have a TUI and 22 markdown files.
How one purpose-built terminal UI and a suite of Claude Code skills - executable, PR-reviewed runbooks - replace a platform team for a six-person startup.
6 min readdev-toolingclaude-codeskillsdeveloper-experienceHow a Tiny Team Ships Like a Big One · The Partner Track
The partner scorecard we generate before every QBR
QBR prep used to be a week of spreadsheet archaeology performed by the most expensive person in the room. We generate the scorecard instead - co-sell pipeline, funding, Marketplace, propensity - and let the humans spend the meeting negotiating.
4 min readawspartnerqbrco-sell-indexproductHow a Tiny Team Ships Like a Big One · The Product Track
One deal, 468 facts, and nothing regenerated twice
Naive AI features re-analyze everything on every change and die of cost and latency. Our deal analysis is event-sourced: changes invalidate only the sections they touch.
4 min readai-engineeringarchitecturecostproductHow a Tiny Team Ships Like a Big One
Claude reviews every PR we open. Only the risky ones ever reach a human.
How we codified review judgment as a severity-by-scope matrix so an AI reviewer can be the only reviewer for routine changes - and never for the risky ones.
5 min readcode-reviewclaudebedrockci-cdHow a Tiny Team Ships Like a Big One · The Partner Track
The difference between being listed on Marketplace and selling through it
One customer was listed on AWS Marketplace for a year with zero transactions. What changed wasn't the listing - it was learning whose money a private offer actually spends.
4 min readawspartnermarketplaceprocurementHow a Tiny Team Ships Like a Big One · The Product Track
How we know the AI is right: dueling judges
Our LLM judge is audited by a second, independent judge on a stronger model - and we compute the agreement statistic before trusting either. On evals that survive contact with production.
4 min readevalsllm-judgeai-engineeringclaudeHow a Tiny Team Ships Like a Big One
The boring rails: how code gets from a laptop to prod here
Path-filtered CI, a release train that cuts itself, migrations as first-class deploy steps, and guardrails with scar tissue - the dull pipeline that makes agent velocity safe.
6 min readci-cdawsreleasesmigrationsHow a Tiny Team Ships Like a Big One · The Partner Track
Marketplace listing as a feature, not a checkbox
Getting listed on AWS Marketplace is a one-time project. Selling through it is a workflow that recurs with every deal - so we built the private-offer path into the product.
4 min readawspartnermarketplaceprivate-offersproductHow a Tiny Team Ships Like a Big One · The Product Track
Which model for which job
One production job took 49 minutes because everything ran on the biggest model. On routing: where frontier models earn their price, where small ones win, and cascades for everything between.
4 min readllmcostai-engineeringclaudeHow a Tiny Team Ships Like a Big One
Dollar budgets, scope guards, and an agent's right to say no
Trust in autonomous agents came to us as a stack of constraints - six plain layers, and the two invoices that taught us to build them.
6 min readagentsguardrailsfinopsclaudebedrocktrustHow a Tiny Team Ships Like a Big One · The Partner Track
Co-sell is a data problem before it's a sales problem
A customer's co-sell program stalled for a year. The fix wasn't sales motivation - it was that nobody could answer 'which of our deals have AWS in them?' from the CRM.
4 min readawspartnerco-selldataHow a Tiny Team Ships Like a Big One · The Product Track
An agent that knows when to stop
Fixed iteration counts are the wrong loop condition for research agents. Ours stop when their evidence is sufficient - and their runtime dropped by half.
4 min readagentsai-engineeringclaudepatternsHow a Tiny Team Ships Like a Big One
100% of our code is written by AI. Here's the system that makes that boring.
From 12% to 100% AI-generated code in six months - and the four-layer operating model of specs, agents, AI review, and human gates that makes it safe.
5 min readai-engineeringagentsclaudeprocessspec-driven-developmentHow a Tiny Team Ships Like a Big One · The Partner Track
One click from opportunity to ACE
Co-sell mostly dies at the swivel chair: a rep re-typing a CRM opportunity into Partner Central. We wired ACE submission into the deal view, with a human deciding what AWS gets to see.
5 min readawspartneraceco-sellproductHow a Tiny Team Ships Like a Big One · The Product Track
Reading an industry's public exhaust
The buying signals enterprises pay for are lying in public - filings, earnings calls, job postings, regulatory actions - in volumes no human team can read. So we stopped using humans.
5 min readai-engineeringsignalsdataproductHow a Tiny Team Ships Like a Big One
The agent that files bugs before they happen
Our daily QA agent reads code instead of logs - and fixed a P1 crash that had zero production occurrences. On prevention, precision, and why a silent bot is a feature.
6 min readagentsqapreventionclaudecode-reviewHow a Tiny Team Ships Like a Big One · The Partner Track
Zero to co-selling with AWS, in the right order
New AWS partners reach for the Marketplace listing and the badge first. Watching alliance teams that work, the right first move is less glamorous: your own deal data.
7 min readawspartnerco-sellalliancesHow a Tiny Team Ships Like a Big One · The Product Track
The 16x problem: your CRM doesn't know who's in the deal
Native automation matched 0.74% of a customer's 425,000 emails to opportunities. On entity resolution - the unglamorous problem underneath every deal-intelligence promise.
5 min readai-engineeringentity-resolutioncrmproductHow a Tiny Team Ships Like a Big One
Anatomy of our SRE fleet: from CloudWatch error to merged fix in 12 minutes
How a fleet of narrow, budgeted Claude agents on AWS turns production exceptions into reviewed, tested pull requests - and why a human still merges every one.
8 min readagentssreawsclaudebedrockHow a Tiny Team Ships Like a Big One
Welcome to Aithon Engineering
Who we are, what we build, and why a team of six builders decided to write an engineering blog.
3 min readmeta