Build Code Nexus Assist

Episode 1

Code Nexus Assist Gets a Real Brief

AI-103 planning and generation: choosing services and regions against a brief, deployments and quotas, project connections, managed identity instead of shared keys, bounded retries, and turning requirements into measurable acceptance criteria.

Episode 2

The Assistant Invented a Refund

AI-103 retrieval and grounding: ingestion (published-only, latest-version, chunking, OCR and enrichment for scanned sources), embeddings, semantic, keyword and hybrid retrieval, chunk-size trade-offs, document-level access control, freshness, and proving an answer is grounded in permitted sources.

Episode 3

Creators Want Visual Superpowers

AI-103 vision and media: reading image and video analysis (text, objects, regions, scene, confidence), writing supported alt text and visual descriptions, video understanding as timed segments, masked edits limited to a region, image generation controls and policy, embedded-text injection, provenance labelling and human review.

Episode 4

Lost in Translation, Found in Tests

AI-103 text and speech: language detection, sentiment, entities, structured JSON output contracts, translation and tone, transcription confidence, speech translation and synthesis as separate stations, audio reasoning and its limits, explicit error handling, regression tests with known-answer fixtures.

Episode 5

Receipts, Screenshots and Voice Notes

AI-103 content extraction: analyzers for documents, images, audio and video, OCR and layout, structured and Markdown output, per-field confidence, evidence-linked records, indexing clean representations, reconciling contradictory readings, and routing uncertainty to review instead of inventing values.

Episode 6

Give the Assistant Tools

AI-103 agents: roles and goals, tool contracts (name, parameters, side effects, approval), function calling, separating user data and tool output from instructions, object-level authorisation inside tools, scoped and per-user memory, per-agent tool grants, supervised multi-agent hand-offs, and traces as evidence.

Episode 7

The Agent Did What?

AI-103 responsible AI, security and operationalisation: trust boundaries and indirect prompt injection, guardrails, scoped tools and per-agent grants, output checks, human approval for consequential actions, auditing through traces, and safety and quality evaluation with attack and benign regression sets and error analysis.

Episode 8

The AI Engineering Review

AI-103 across all domains: planning and quotas, retrieval and search health, vision, language and speech, extraction, agents and memory, and security — plus monitoring for quality, latency, cost and drift, held-out evaluation, prompt and configuration tuning, failure analysis, staged release, and recording model, index and prompt versions.