This catalog contains only the 39 HUMAI articles that completed the independent recovery audit on 17 July 2026. It omits the remaining legacy archive until each item receives the same editorial, sourcing, reuse, metadata, and publishing checks.
Choose the reader task closest to your current decision. Each description explains the question the linked article is designed to answer.
Choose a model, tool, or device
- Google Antigravity vs Cursor in 2026: compare product boundaries, agent behavior, plans, privacy, and security.
- Nano Banana vs Nano Banana Pro: separate the documented image-model variants and their available controls.
- TurboLearn is now Turbo AI: understand the rebrand, current features, plan ambiguity, privacy, and alternatives.
- Tavily vs Exa vs Perplexity vs You.com: choose a search API by retrieval task, pricing unit, SDK, and data boundary.
- Four AI gadgets in 2026: compare purchase cost, subscription burden, privacy, and practical fit.
- Seven AI assistants compared: route work by product surface, plan boundary, data control, and permission risk.
- Runway review: examine models, credits, exports, privacy, rights, and a controlled trial protocol.
- Cursor vs Antigravity vs Claude Code vs Windsurf: compare coding surfaces, agent permissions, plans, and data handling.
- Perplexity review: decide when cited answers help and how to verify them before reuse.
- Free AI chatbots for work: test zero-cost plans against a fixed task set without relying on rankings.
- Fireflies.ai review: map consent, recording routes, retained artifacts, storage, and plan limits.
- VisualGPT buyer check: review credits, renewal terms, refunds, privacy, deletion, and output-rights questions.
- v0 by Vercel: follow a prompt-to-app handoff through code custody, Git, deployment, and production review.
Build software and controlled workflows
- Vibe coding in 2026: use a build process with scoped changes, tests, review, secrets protection, and rollback.
- AI search optimization for SaaS: improve crawlability, evidence, structured pages, and measurement on Google.
- Headless CMS for designers: decide whether content reuse and channel needs justify the added system complexity.
- Build an agent in Microsoft Copilot Studio: create a draft-only workflow with permissions, tests, approval, and incident handling.
- Create a GPT in ChatGPT: define instructions, knowledge, actions, evaluation, privacy, and release controls.
- Real-time AI inference: interpret latency budgets across model, network, queue, and application stages.
- Claude Code for designers: move from Figma context to reviewed code with checkpoints and visual comparison.
- Claude Haiku 3.5 to 4.5 migration: update model IDs, parameters, tools, prompts, tests, limits, and data controls.
- Gemini File Search lifecycle: operate stores, ingestion, retrieval, citations, deletion, pricing, and logs.
Manage work, policy, and business risk
- Write a workplace AI policy: adapt a practical template for approved uses, data, review, vendors, and incidents.
- Read AI job-market evidence: distinguish exposure, task change, hiring signals, forecasts, and measured outcomes.
- Can AI enable a 15-hour workweek?: calculate whole-week savings and include review, maintenance, quality, and coverage.
- Audit an AI and SaaS stack: measure total cost, overlap, permissions, portability, and exit paths.
- AI automation for freelancers: select one repeatable workflow, control client data, and measure full delivery cost.
- Freelance prompt engineering: frame a service around evaluated business outputs instead of selling isolated prompts.
- Walmart AI status map: separate live customer features, internal operations, pilots, announcements, and missing evidence.
- When not to use AI: apply stop rules for sensitive data, emergencies, accountability, accessibility, and weak economics.
Understand models, infrastructure, and public evidence
- How ChatGPT works: trace a prompt through tokens, context, prediction, tools, and safety layers.
- AI reasoning models: learn what test-time compute changes and where evaluation still matters.
- On-device AI: identify which tasks can run locally and what hardware, privacy, and update limits remain.
- Apple M5 vs NVIDIA B200 vs Google TPU7x: choose among local, owned, and rented compute routes without false benchmark equivalence.
- LatamGPT 1.0 status: inspect the public model artifact, language scope, governance record, and disclosure gaps.
Create and process visual media
- Figma AI vs Google Stitch: match each product to interface work, handoff needs, access, and review.
- Google Pomelli review: understand its website-to-campaign workflow, input needs, editing, access, and privacy.
- Free image upscaling: compare four routes with one reader-run test and consistent source images.
- Five video upscaling routes: design a fair local, editor, and cloud comparison using retained artifacts.
What inclusion means
Inclusion means the article passed the July 2026 recovery checks. It does not mean every linked product is endorsed, every plan detail will remain current, or one answer fits every reader. Check the article's source links and date-sensitive notes before making a purchase or production decision.