DROPGoogle launches Gemini 3.8 Flash TTS and Flash-Lite TTS///DROPGemini Omni brings 1080p generation to Google Vids///BREAKClaude Opus 5.5 is here: 1M context, $4 input / $20 output///BREAKOpenAI adds GPT-6 Sol and Luna to the GPT-6 family///PAPEROpus 5.5's system card documents gains—and evaluation limits///BREAKGrok 4.7 launches at $2 input / $6 output per million tokens///DROPGrok Voice Transcribe 2.0 targets difficult real-world audio///TRENDGoogle steers new Gemini API projects to 3.8 Flash or 3.5 Flash-Lite///DROPGoogle launches Gemini 3.8 Flash TTS and Flash-Lite TTS///DROPGemini Omni brings 1080p generation to Google Vids///BREAKClaude Opus 5.5 is here: 1M context, $4 input / $20 output///BREAKOpenAI adds GPT-6 Sol and Luna to the GPT-6 family///PAPEROpus 5.5's system card documents gains—and evaluation limits///BREAKGrok 4.7 launches at $2 input / $6 output per million tokens///DROPGrok Voice Transcribe 2.0 targets difficult real-world audio///TRENDGoogle steers new Gemini API projects to 3.8 Flash or 3.5 Flash-Lite///
DARC.DEV

[ 02 — 35 available field guides ]

Learn something. Use it today.

Current release briefings and durable engineering skills. Every guide connects the explanation to a practical procedure, checklist, source and next step.

35 field guides · practical sections · checklists · source links

Editorial illustration for this AI field guideStart

Analysis · 4 min

Opus 5.5. GPT-6. Grok 4.7. The model map just changed.

A source-checked field guide to the September releases: what is actually available, what it costs, and what needs changing in your stack.

2 sections · 11 sources
Editorial illustration for this AI field guideBuild

Guide · 3 min

Moving to Opus 5.5? Read the API changes before the benchmarks.

Always-on thinking, cheaper cache reads and safeguard fallback routing: the operational details behind Anthropic's 22 September release.

2 sections · 4 sources
Editorial illustration for this AI field guideBuild

Comparison · 3 min

GPT-6 Astra, Sol and Luna: three tiers, not one default answer.

The current API IDs, 1.05M context windows and price conditions—plus the migration constraints that a model leaderboard will not tell you.

2 sections · 6 sources
Editorial illustration for this AI field guideStart

Guide · 3 min

Gemini 3.8 is a family. Flash, Live, TTS and Cyber do different jobs.

A practical map of Google's September models, with real availability dates, introductory pricing and the distinction between generation and live dialogue.

2 sections · 5 sources
53
Analysis
Build

Analysis · 3 min

Grok 4.7: what the new model actually changes for builders.

500K context, a $2/$6 standard token rate, and published long-horizon coding results—with the evaluation conditions still attached.

2 sections · 2 sources
Editorial illustration for this AI field guideBuild

Methodology · 3 min

How to benchmark coding agents without gaming the result

Design a reproducible evaluation on a real repository: task cards, protected acceptance tests, fixed budgets and cost per reviewed change.

2 sections · 4 sources
28
Comparison
Start

Comparison · 3 min

Claude, GPT, Gemini or Grok? Choose by the job, not the brand.

A selection framework for Opus 5.5, GPT-6, Gemini 3.8 Flash and Grok 4.7: hard constraints first, representative tests second, operational cost third.

2 sections · 6 sources
98
Analysis
Deep

Analysis · 3 min

Open weights now: DeepSeek V4.1, Qwen3.8 and Mistral's active lineup

What the official repositories and changelogs actually say about versions, context, deployment and licenses.

2 sections · 8 sources
47
Tutorial
Build

Tutorial · 3 min

Prompt caching, properly measured: prefixes, writes, reads and real cost

A worked cost example, current GPT-6 cache behavior and a debugging checklist for stable prefixes, cache boundaries and reuse.

2 sections · 6 sources
Editorial illustration for this AI field guideBuild

Guide · 3 min

RAG is not dead. Your context needs a design.

Retrieval, freshness, source authority and evaluation belong in one system. Design the context policy, then diagnose the first failing stage.

2 sections · 5 sources
Editorial illustration for this AI field guideBuild

Tutorial · 3 min

Local AI: plan the memory before downloading the weights

Quantization, KV cache, CPU offload and hardware costs. What a 24GB GPU can—and cannot—promise for large models.

2 sections · 2 sources
28
Guide
Start

Guide · 3 min

Build your first agent around a small, inspectable loop

Plan, act, observe, stop. Use current model APIs without surrendering budgets, logs or human approval.

2 sections · 5 sources
10
Guide
Deep

Guide · 3 min

Evals: separate a published score from your product's reliability

Benchmark version, harness, effort and uncertainty: the minimum evidence needed before turning a number into a deployment decision.

2 sections · 6 sources
Editorial illustration for this AI field guideStart

Comparison · 3 min

AI video in September: track the model, the product and the rollout

A production checklist built around current Gemini Omni, TTS and creator-tool workflows: usable shots, continuity, rights, narration and delivery.

2 sections · 5 sources
68
Guide
Deep

Guide · 3 min

The agent security checklist still matters after the model upgrade

Narrow tools, output filtering, approval gates and incident traces—read alongside the new model system cards.

2 sections · 5 sources
50
Workflow
Build

Workflow · 3 min

A maintainable Claude Code workflow after the Opus 5.5 release

Repository instructions, planning gates and reviewable diffs—plus the model settings that now need a fresh check.

2 sections · 5 sources
65
Methodology
Build

Methodology · 3 min

Measure AI-search citations: mentions, evidence and actual visits

A reproducible citation-tracking plan with distinct outcome categories, stable queries, evidence records and honest attribution limits.

2 sections · 3 sources
52
Methodology
Build

Methodology · 3 min

Design a model router around quality, cost and current prices

Routing, context budgets and caching as a measurable system. Compare accepted-task cost while preserving a fixed quality threshold.

2 sections · 6 sources
Editorial illustration for this AI field guideStart

Foundations · 4 min

Tokens, context and tools: the mental model that makes AI less mysterious

Understand what a model sees, what it can do, and why a million-token window is not the same as memory. A practical starting point before you choose a stack.

4 sections · 4 sources
10
Tutorial
Build

Tutorial · 3 min

Structured output is a contract—not a truth machine

Design extraction and classification pipelines that survive missing fields, refusals, bad evidence and provider errors. Includes a concrete validation checklist.

4 sections · 4 sources
58
Guide
Deep

Guide · 3 min

MCP in 2026: connect tools without giving away the keys

The July 28 specification, HTTP authorization, token audience checks and the practical difference between a tool description and permission to execute.

4 sections · 4 sources
69
Tutorial
Build

Tutorial · 3 min

Put prompt changes behind a test gate, not a vibes check

A practical promptfoo workflow: fixtures, deterministic assertions, calibrated judges, budget caps and CI artifacts you can actually debug.

4 sections · 3 sources
Editorial illustration for this AI field guideBuild

Guide · 3 min

Hybrid search: when exact words matter as much as meaning

Build a retrieval test set before picking pgvector or Qdrant. Learn what to measure, where permission filters belong and when fusion actually helps.

4 sections · 3 sources
10
Guide
Build

Guide · 3 min

Trace your AI application without creating a second data leak

A minimal observability plan with useful spans, version metadata, cost accounting and redaction before export.

4 sections · 4 sources
24
Workflow
Start

Workflow · 3 min

An AI workflow that drafts, validates and waits for approval

Build a safer automation pattern for incoming requests: deterministic routing around a narrowly scoped model step, with idempotent writes and a visible error queue.

4 sections · 4 sources
21
Guide
Deep

Guide · 3 min

Your agent can resume. Can it avoid doing the same thing twice?

Durable execution, checkpoints and external side effects explained through an approval-and-resume workflow.

4 sections · 3 sources
67
Workflow
Start

Workflow · 3 min

A better research notebook starts with better sources

A practical Gemini Notebook workflow for a cited brief, a disagreement map and a study plan—without outsourcing source judgment.

4 sections · 2 sources
60
Guide
Start

Guide · 3 min

Research with AI: turn a list of links into an evidence trail

A source-led workflow for web research, provider announcements and technical comparisons. Includes a claim ledger and a practical verification rubric.

4 sections · 4 sources
Editorial illustration for this AI field guideBuild

Guide · 3 min

Your voice agent sounds human. Does it handle interruptions and numbers?

Evaluate the entire speech-to-action loop: recognition, reasoning, turn-taking, tool completion, consent and latency.

4 sections · 5 sources
43
Workflow
Start

Workflow · 3 min

From generated image to publishable asset: the checks that matter

A practical workflow for references, versioning, editing, typography, privacy and rights—with current Midjourney and Gemini examples.

4 sections · 4 sources
17
Guide
Build

Guide · 3 min

The preview works. Here is what to check before real users arrive.

A launch checklist for apps built with v0, Lovable, Replit or a coding agent: authorization, persistence, secrets, failure states and rollback.

4 sections · 4 sources
48
Guide
Build

Guide · 3 min

Agent permissions: distinguish instructions, approval and isolation

Why a CLAUDE.md rule, a confirmation dialog and a sandbox are different controls—and how to test that they work together.

4 sections · 5 sources
47
Guide
Deep

Guide · 3 min

A model gateway is not a budget until it can stop a request

Virtual keys, database-backed spend tracking, fallback policy and a simple test for real enforcement.

4 sections · 3 sources
Editorial illustration for this AI field guideBuild

Checklist · 3 min

Local AI without hidden cloud dependencies

A practical deployment audit for Ollama, LM Studio and Open WebUI: model weights, embeddings, network tools, API exposure and maintenance.

4 sections · 4 sources
90
Workflow
Start

Workflow · 3 min

Turn a voice note into a useful product brief—not invented certainty

A transcript-to-spec workflow that preserves unresolved questions, distinguishes requirements from ideas and produces testable acceptance criteria.

4 sections · 3 sources