Explore the apidays talks, the From Buzzwords to Blueprint class and the author. Choose an example to edit and ask, or write your own question.
Open a topic to explore its terms and example questions.
apidays Australia 2026 talks
Notes from 18 talks · 28–29 October, Melbourne
01Agents, MCP and events+
How agents react to changes, gather incident evidence, and leave engineers in control.
- Model Context Protocol (MCP)
- OODA: Observe, Orient, Decide, Act
- Outbox pattern
- System 1 router
- Event lineage
- Principal Agent
- Evidence ledger
- Time to context
Try a question
02Agent-ready APIs and identity+
Make APIs discoverable, clear to call, safe to retry, and traceable to the agent and user.
- OpenAPI
- The six D’s of agent readiness
- Semantic descriptions
- Machine-readable schemas
- Delegated identity
- Least privilege
- Idempotency keys
- Retry guidance
- End-to-end tracing
Try a question
03API monetization and partners+
Who pays for agent calls, how permission levels change pricing, and how partner value is measured.
- Read, Recommend, Prepare, Execute
- Wallets, commitments and allowances
- Metering
- Co-sell
- Committed drawdown
- Value per call
- Time to first paid call
- Accountability share
Try a question
04Knowledge graphs and workflow contracts+
Give agents shared meaning, controlled data access, and an explicit order for API calls.
- Ontology
- Subject–predicate–object triples
- SHACL shapes
- PROV-O lineage
- GraphQL Data Graph
- Schema introspection
- Trusted Documents
- Federation
- Arazzo
- OpenAPI and AsyncAPI
- Executable workflow contracts
Try a question
05Agent security and governance+
Control what enters the model, which actions can run, and what answers can leave.
- Prompt injection
- Narrative attacks
- Input and dialog rails
- Retrieval rails
- Execution and output rails
- Scoped credentials
- Human approval
- API execution layer
- LLM gateway
- Token limits and audit records
Try a question
06Harnesses, smaller models and prototypes+
Turn feedback into repeatable checks, choose model trade-offs, and build prototypes with deterministic tools.
- Backpressure
- Sensors and gates
- Deterministic and inferential checks
- Protective friction
- OpenTelemetry
- Agent eXperience Interface (AXI)
- Executable attention
- Skills, CLI and tools
- XML contracts
- Token cost
Try a question
07Delivery, quality and measurement+
Find where work waits, plan quality checks, and decide how to measure a feature before building it.
- Theory of Constraints
- Work in progress (WIP)
- Review queues
- Plan, generate, judge
- Playwright test agents
- CI quality checks
- Experiment gates
- Exposure logging
- Cohort balance
- Conversion rates
- Statistical significance
Try a question
08Architecture and durable execution+
Set domain boundaries and keep long-running business workflows recoverable after failures.
- Modular monolith
- Domain boundaries
- Choreography and orchestration
- BIAN
- ISO 20022
- Temporal
- Workflows and Activities
- Event history
- Durable timers
- Signals
- Sagas and compensation
Try a question
From Buzzwords to Blueprint class
Online class · 20–21 November · 17 concepts and how this bot was built · Register for the class
01The agent itself+
What a model, its tools and its loop do, and where the harness and the context come from.
- tool calling
- AI agent
- agentic AI
- MCP
- harness
- context engineering
- Agent, tool or skill
Try a question
02Control+
Stop untrusted text from giving orders, check actions and answers, and keep evidence of what happened.
- prompt injection
- guardrails
- human in the loop
- observability
- LLM gateway
Try a question
03Many agents and long flows+
Split work between agents, coordinate it with events or a coordinator, and recover long flows after failures.
- multi-agent
- orchestration
- choreography
- event-driven
- durable execution
- agent-ready API
Try a question
04How this bot was built+
The architecture of this bot, the rules that plain code enforces, and what testing changed.
- Defined corpus
- Plain code for fixed rules
- Answer cache
- Citations
- Fallback behavior
- Tests and evaluation
Try a question
About the author
Foyzul Karim · Melbourne
01Foyzul Karim+
Who wrote these notes and taught the class, what the work covers, and where to follow or contact the author.
- Software engineer and architect
- Teacher and public speaker
- Distributed systems
- Node.js platforms
- Cloud architecture
- AI-assisted development
- Courses and videos
- Contact
Try a question