The Kit

Best AI coding tools

1. Cursor— the AI-native code editor (a VS Code fork with the model wired into every surface). Reach for it when you live in an editor all day and want codebase-aware chat, inline edits, and multi-file changes without leaving the keyboard. Caveat: it's an editor switch, not an add-on — the value comes after the muscle-memory tax.
2. Claude Code— Anthropic's agentic coding tool: give it a task in the terminal, it reads the repo, edits files, runs tests, iterates. Reach for it when the job is a task ("fix this bug," "add this endpoint"), not keystroke assistance — it's delegation, not autocomplete. Caveat: agentic sessions consume tokens accordingly; review the diffs like you'd review a junior's PR.
3. GitHub Copilot— the default: autocomplete, chat, and an agent mode inside VS Code, JetBrains, and github.com. Reach for it when you want the lowest-friction start, team seats under existing GitHub billing, or you're not ready to change editors. Caveat: master of no single mode — specialists beat it at their specialty.
4. Windsurf— the other AI-native editor, with an emphasis on agentic "flows" and a gentler price. Reach for it as the Cursor alternative if its flow model fits your style better — try both for a week; this choice is taste. Caveat: the category leader's ecosystem and iteration pace are Cursor's; you're betting on the challenger.
5. Replit— browser IDE plus deployment plus an app-building agent. Reach for it when you want idea-to-deployed-URL in one place — prototypes, internal tools, learners. Caveat: production engineering teams will hit its ceilings; that's not who it's for.

Best AI writing tools

1. Claude / ChatGPT (the frontier chats)— the honest first answer: for most writing, drafting, and editing, a frontier general model beats the specialized wrappers. Reach for one before paying for anything below; the $20 tier covers most writers. Caveat: generic by default — the quality ceiling is your prompting and your taste.
2. Grammarly— grammar, tone, and clarity checking everywhere you type (browser, docs, email). Reach for it when you want always-on proofreading across surfaces rather than a writing destination. Caveat: its AI-writing features trail the frontier chats; you're paying for ubiquity, not brains.
3. Jasper— marketing-content platform: brand-voice profiles, campaign templates, team workflow. Reach for it when a team produces branded marketing copy at volume and needs consistency, not when an individual needs a writer. Caveat: expensive relative to a frontier chat; the premium is the workflow, not the words.
4. Sudowrite— built for fiction: prose continuation, description expansion, story-bible tools. Reach for it if you write novels and want a collaborator that thinks in scenes and chapters. Caveat: niche by design; useless for business writing.

Best AI image generation

1. Midjourney— still the aesthetic benchmark: the most opinionated, art-directed default look in the field. Reach for it for concept art, mood, and anything where beauty is the deliverable. Caveat: the house style is strong — fighting it is work — and precise text-in-image and layout control trail the field.
2. Ideogram— the text-rendering specialist. Reach for it for anything with words in the image: posters, logos mockups, memes, packaging comps. Caveat: overall aesthetic ceiling sits below Midjourney's for pure imagery.
3. Flux (Black Forest Labs)— the open-weight frontier: run it locally or via API, fine-tune it, build on it. Reach for it when you need control, reproducibility, or self-hosting — the tinkerer's and pipeline-builder's choice. Caveat: you're assembling your own workflow; there's no polished consumer app in the box.
4. DALL·E / GPT image generation— image gen inside ChatGPT, with conversational editing. Reach for it when convenience wins: you're already in the chat, and iterating by talking ("same image, but at night") is the feature. Caveat: raw output quality is contested territory every quarter; convenience, not supremacy, is the pitch.
5. Adobe Firefly— generation trained on licensed content, integrated into Photoshop and the Creative Cloud. Reach for it in commercial pipelines where rights provenance and indemnification matter, and for generative fill inside real design work. Caveat: raw generation quality trails the leaders; the moat is legal comfort plus Photoshop.

Best AI video generation

The most volatile category in the Kit — re-rank quarterly.

1. Runway— the production-workflow leader: generation plus editing tools, camera controls, and motion direction aimed at people who ship video. Reach for it when you need control over the shot, not just a pretty clip. Caveat: credits burn fast; budget for iteration.
2. Sora (OpenAI)— the headline generator, strong at coherent scenes and physics-ish motion. Reach for it for cinematic single shots and ideation inside the OpenAI ecosystem. Caveat: access tiers and usage limits shift; production pipelines need more edit control than it offers.
3. Google Veo— the frontier competitor with strong realism and native audio generation. Reach for it for realistic footage with sound, and if you're already in the Google/Gemini ecosystem. Caveat: same class of limits — prompt-and-pray beats shot-directing for now.
4. Kling— the volume value play: competitive quality at aggressive pricing. Reach for it for high-iteration workflows where cost per attempt matters more than peak fidelity. Caveat: consistency across generations is the trade.
5. Pika— the fast, playful option with effect-driven templates. Reach for it for social clips and quick stylized shots. Caveat: aimed at fun over film; ceiling is lower by design.

Best AI search and answer engines

1. Perplexity— the category-defining answer engine: cited answers, follow-up threads, focus modes. Reach for it when you'd otherwise open ten tabs — research questions with checkable sources. Caveat: citations reduce hallucination, they don't eliminate it — click through before you repeat a claim. (The whole category note: verify before you cite.)
2. ChatGPT (search)— web-grounded answers inside the default AI app. Reach for it when the question is one turn of a longer working conversation rather than a research session. Caveat: search quality is the feature, not the product; dedicated engines cite more carefully.
3. Google AI Mode / AI Overviews — the incumbent's answer, fused with the live index and free at Google scale. Reach for it for fresh, local, and commercial queries where the index matters more than the reasoning. Caveat: the ad-funded elephant — the layer's clearest conflict between answer quality and business model.
4. Kagi— paid, ad-free search with AI answers on top. Reach for it if you want search whose customer is you, not advertisers — the alignment argument in product form. Caveat: paying for search is a philosophical purchase; the AI features are solid, not frontier.

Best AI meeting notes

1. Granola— the notes-first darling: it enhances your rough notes with the transcript instead of replacing you, and no bot joins the call. Reach for it if you take notes anyway and want them upgraded, not automated. Caveat: desktop-centric; check platform coverage for your setup.
2. Fireflies— the workflow workhorse: transcription, summaries, action items, CRM sync, across every major meeting platform. Reach for it when the pipeline matters — meetings in, structured data out, into your existing tools. Caveat: a bot joins your calls; some rooms mind.
3. Otter— the transcription veteran with live transcripts and a generous free tier. Reach for it for straightforward transcribe-and-summarize at the lowest cost of entry. Caveat: the AI-assistant layer is thinner than the newer entrants'.
4. Zoom / Teams / Meet built-ins — the incumbents' free-with-your-seat summaries. Reach for them when IT already approved them and "good enough" is the bar. Caveat: the classic incumbent trade — placement over quality; summaries trail the specialists.

Best agents & automation

1. Zapier— the automation incumbent, now with AI steps and agent workflows across thousands of app integrations. Reach for it when the job is connecting SaaS tools with some AI judgment in the middle — the integration catalog is the moat. Caveat: per-task pricing compounds; complex flows get expensive and brittle.
2. Make— the visual automation power tool: cheaper per operation, more control, steeper learning curve. Reach for it when Zapier's pricing stings and you enjoy a node graph. Caveat: you will become the maintainer of your own spaghetti.
3. n8n— open-source workflow automation, self-hostable, with first-class AI/agent nodes. Reach for it when you want your automation on your infrastructure with no per-task meter — the engineer's choice of the three. Caveat: self-hosting is a responsibility, not a feature, until you've been paged for it.
4. Lindy— AI-agent-as-employee: build assistants that handle email, scheduling, CRM updates from natural-language instructions. Reach for it to delegate a defined recurring task rather than build a workflow by hand. Caveat: the demo-to-dependable gap is the whole agent category's open wound; supervise before you trust.
5. Claude Code / computer-use agents— general agentic tools pointed at automation: file wrangling, report generation, browser tasks. Reach for them for ad-hoc, one-off automations that don't merit a standing workflow. Caveat: per-run token cost and nondeterminism make them wrong for high-frequency pipelines.

Best vertical AI (the specialist shelf)

Different rules apply here: these are enterprise sales, not credit-card signups — "reach for it" means "shortlist it."

1. Harvey— the legal-AI flagship: research, drafting, and review built for law-firm workflow, sold firm-wide. Shortlist it if you're a firm or in-house team buying a platform, not a tool. Caveat: platform pricing and rollout; solo practitioners look downlist.
2. Spellbook— contract drafting and review inside Microsoft Word, priced for smaller teams. Shortlist it if contracts are the job and Word is the habitat. Caveat: contracts-focused — it's not a research platform.
3. Abridge— ambient clinical documentation: the visit is recorded, the note writes itself into the EHR. Shortlist it for health systems fighting clinician burnout; this category has some of the clearest ROI evidence in vertical AI ⚑ unverified. Caveat: enterprise integration project, measured in quarters.
4. Sierra— agentic customer service that resolves tickets end-to-end, priced per resolution. Shortlist it when support volume is the P&L problem and outcome-based pricing appeals. Caveat: outcome pricing means auditing what "resolved" means in the contract.
5. Hebbia— document intelligence for finance: analyze data rooms, filings, and research at diligence scale. Shortlist it for funds and banks where the workload is reading at volume. Caveat: finance-shaped pricing; this shelf is not for browsing.