# Vibebrator

> The vibe coding workbench: 20 tested prompts you can engineer, a 60-term
> dictionary, and 95 curated tools. Everything below is the full content of
> the site in plain text. Nothing is hidden behind JavaScript.

Site: https://www.vibebrator.com
Updated: 2026-09-02

## The Prompt Deck

### The Titan Audit  (audit)
Holds every file to the standard of the most valuable companies on earth. Each quality maps to a testable property, every finding gets a P0–P3 severity and a concrete fix.
When: Before you ship something you will have to maintain. Or when a build works and you cannot tell whether it is any good.

### Lock the Stack  (build)
Pins the framework, styling and data layer before a line is written, and forbids silent substitutions.
When: First prompt of any new build.

### Security Boundary  (build)
States the multi-tenant boundary explicitly and forces enforcement into the database rather than application code.
When: Any app where two customers' data must never meet.

### Plan Before It Types  (build)
Forces a research and planning pass before any edit, and makes the agent name what it is unsure about.
When: Unfamiliar codebase, or any change touching more than one file.

### Define Done  (ship)
Converts a vague finish line into observable, demonstrable criteria the agent has to evidence.
When: Before you accept any piece of work.

### Hand It An Error  (debug)
Gives the model the trace plus what you expected, what changed, and what you already ruled out — then demands root cause before fix.
When: Every single time something breaks.

### Policy Recursion  (debug)
Explains why a row-level-security policy that queries its own table recurses, and fixes it without weakening isolation.
When: Postgres 42P17, or any policy that references its own table.

### Verify The Imports  (debug)
Forces every imported symbol to be confirmed against your installed version before you read a line of implementation.
When: First 30 seconds after any generated file appears.

### Stop Refactoring  (refine)
Bounds the blast radius to one file and one function, and gives the model a legitimate place to put its other opinions.
When: Any edit to code that already works.

### Stop It Agreeing  (refine)
Requires the strongest case against your approach, and the most likely failure mode, before any implementation.
When: Before any decision that is expensive to reverse.

### Show Your Work  (ship)
Demands the diff, the raw test output, the commit SHA, and an explicit list of what could not be verified.
When: Every time an agent claims it finished.

### Hunt Regressions  (ship)
Turns the model against its own diff, looking for unrequested behaviour changes and swallowed error paths.
When: After the fix, before the merge.

### Build Receipt  (ship)
Ends a session with a signed receipt: files touched, commands run, what was verified, what was not. The artifact you paste into the PR.
When: The last message of every working session.

### Reversible Migration  (build)
Forces a database migration to arrive with a down migration, a data-safety statement, and a dry run — before it touches anything.
When: Any change to a table that already has rows in it.

### Explain It Like I Own It  (refine)
Turns a file you were handed into something you could defend in a meeting: what it does, what would break it, and what you would need to change first.
When: Before you accept ownership of anything an agent — or a contractor — produced.

### Red Before Green  (build)
Makes the agent write a failing test that pins the behaviour, run it, show the red, and only then implement.
When: Any feature where "it works" would otherwise be your only evidence.

### Write the Rules File  (build)
Has the agent read the repository and draft the standing rules file — stack, commands, conventions, forbidden moves — from what is actually there.
When: Day one of any repo an agent will work in more than once.

### Measure Before Optimising  (debug)
Refuses performance work until there is a number, a reproduction, and a target — then bounds the change to the measured hot path.
When: Anything is "slow" and someone is about to start rewriting.

### Cold Handoff  (refine)
Produces the message you paste at the top of the next session: state, decisions already made, what not to redo, what is next.
When: Context is about to run out, or you are switching models.

### Dependency Interrogation  (audit)
Audits the dependency list: does it exist, is it maintained, and why is it here. Catches hallucinated packages and dead weight in one pass.
When: After any session that touched package.json, requirements.txt, or the lockfile.

## The Confidence Circuit (lessons)

### 1. Read the diff before you run it  (25 min)
Install the habit of reading every generated change before executing it, and learn what to look for when you do.
Check: You can name, from memory, one hunk in that diff you would have shipped without noticing if you had not read it. If you cannot, the diff was too small to learn on — pick a bigger one.

### 2. Write the rules file  (40 min)
Produce an AGENTS.md / CLAUDE.md for a real repo, derived from what is actually in it, and prove a fresh session can work from it alone.
Check: The fresh session runs the test suite correctly without asking you a single question. If it asks, the answer goes into the rules file and you try again.

### 3. Red before green  (45 min)
Ship one feature where the failing test existed, and was seen to fail, before the implementation did.
Check: You have a screenshot or a saved log of the same test red, then green, and you ran the green one yourself. Keep both in the pull request.

### 4. Know your building  (30 min)
Map your own app onto the ten layers of the stack, and locate your last three bugs on that map.
Check: All three bugs are on a floor, and at least one of them is on a different floor than you first assumed. If all three landed exactly where you guessed, you were already doing this — move on.

### 5. Lock the doors  (45 min)
Close the three ship-blockers in the Vibe Check on a real project: no secrets in history, a test proving isolation, and an owner check on every route.
Check: All three Vibe Check blockers are a yes, and you can point to the artifact behind each one: the scan output, the isolation test, the route table.

### 6. Ship with a receipt  (20 min)
End a session with a Build Receipt and a Cold Handoff, and use them — not your memory — to start the next one.
Check: The next session's first action matched the handoff, and nothing from the receipt's Changed list was redone. Two sessions, one thread of work, no memory required.

## The Stack (your app as a building)

### 1. DNS — the sign on the highway
Turns the name into an address. It is the only layer you rent instead of own.
What bites: A wrong record makes the whole building vanish while every server is fine — and the fix takes as long as the old record’s TTL says it takes.

### 2. CDN / Edge — the nearest doors
Copies of the lobby placed close to every visitor, so the page opens fast wherever they are.
What bites: Stale copies. You shipped the fix; the door nearest your visitor is still handing out yesterday’s build. Purge, then check from somewhere that is not your own machine.

### 3. HTTPS / TLS — the bag check
Proves the door belongs to you and seals everything said inside.
What bites: A certificate issued for one name and served on another. The browser calls it a hostname mismatch; your visitor calls it a scam warning.

### 4. Frontend — the lobby and the map
Everything the visitor sees and touches. It runs on their device, not yours.
What bites: Anything the browser can render, the visitor can read. Secrets, prices and rules that live only here are suggestions.

### 5. Auth — the membership desk
Works out who is at the door. That is authentication, and it is all this desk does.
What bites: The desk checks ID. It does not decide which floors you may enter — that is authorization, and it has to live on every hallway. Confusing the two is how login loops and open doors happen.

### 6. API routes — the hallways
Named paths from the lobby to the stores. Every request walks one.
What bites: A hallway without its own badge check is an open door, whatever the lobby says. Validate at the entrance of every route, not once at the front.

### 7. Services — the stores
Where the work happens: pricing, bookings, invoices. The business logic.
What bites: An agent will happily put store logic in the lobby. A rule that exists in two places drifts, and nobody notices until the two disagree.

### 8. Third-party — the food court
Other people’s businesses leased inside your building: payments, email, maps.
What bites: Their outage looks like yours. Their webhooks arrive unsigned unless you check. Their keys, leaked once, run up your bill.

### 9. Background jobs — the loading dock
Work that happens off the sales floor: sending, exporting, retrying, cleaning up.
What bites: Jobs that fail in silence. If nothing counts the failures and retries them, the dock fills up and nobody on the floor can tell.

### 10. Database — the stockroom
Everything you actually own, in one place, behind one door.
What bites: The lock has to be in the elevator, not on the lobby sign — that is row-level security. And a backup you have never restored is a hope, not a backup.

## The Lexicon

### agent memory  [EY-juhnt MEM-uh-ree]  noun
State an agent carries across turns or sessions: recent history plus durable notes.
Distinct from the context window, which is what fits right now. Memory is what gets chosen to come back.

### agentic coding  [ey-JEN-tik KOH-ding]  noun
Development in which an AI agent runs a loop of gathering context, acting with tools, and verifying results, rather than answering one prompt at a time.
The grown-up sibling of vibe coding. The distinguishing feature is the verify step: an agent that cannot check its own work is just a very fast typist.

### AI slop  [ey-ahy slop]  noun
Generated output that is voluminous, generic, and low in value.
Merriam-Webster made slop its 2025 Word of the Year, the same year Collins chose vibe coding, which is a fairly complete summary of how that year went. The term now covers code, pull requests, documentation, and issues.

### blast radius  [blast REY-dee-uhs]  noun
The set of files, systems, and behaviors a change can affect.
State it in the prompt and the agent stays in its lane. Leave it implicit and you get a one-line fix wrapped in four hundred lines of unrequested tidying.

### brownfield  [BROWN-feeld]  noun
An existing codebase with behavior that must be preserved while it is changed.
The real market and the hard problem. Most software work is brownfield, and most vibe-coding content pretends otherwise.

### cargo culting  [KAHR-goh KUHLT-ing]  noun
Copying the visible form of a practice without the understanding that makes it work.
From Richard Feynman's 1974 Caltech address on cargo cult science. The modern version is a repository full of configuration files copied from a blog post, none of which anyone can explain.

### chain of thought  [cheyn uhv thawt]  noun
A model working through intermediate reasoning steps before producing an answer.
It improves accuracy on multi-step problems. It is not a transcript of the model's actual internals, and reading it as one is a category error.

### characterization test  [kar-ik-ter-uh-ZEY-shuhn test]  noun
A test written to capture what existing code actually does, rather than what it should do.
The essential tool for brownfield work. Before you let an agent near legacy code, you pin its current behavior, because otherwise you cannot tell a fix from a regression.

### confident wrongness  [KON-fi-duhnt RAWNG-nis]  noun
Output that is fluent, plausible, well-formatted, and incorrect.
The signature failure mode of the whole field. Human writing signals uncertainty through hedging and mess; generated writing does not, so fluency stops being evidence of anything.

### constraint block  [kuhn-STREYNT blok]  noun
The section of a prompt that bounds the solution space: stack, forbidden actions, invariants, and things the model must not touch.
The highest-leverage part of any prompt and the part most often skipped. A model will not infer your security boundary; if isolation is not in the constraints, it does not exist in the output.

### context engineering  [KON-tekst en-juh-NEER-ing]  noun
The practice of deliberately assembling what enters a model's context window: retrieval, memory, system instructions, tool definitions, and prior turns.
It displaced prompt engineering as the term of art, and the shift matters. The leverage moved from how you phrase the request to what you put in front of the model, and just as importantly, what you keep out.

### context explosion  [KON-tekst ik-SPLOH-zhuhn]  noun
The degradation in output quality that occurs when an agent reasons over too much material at once.
More context is not more understanding. Past a point the signal you needed is buried in the material you added, and the usual defense is a subagent with a narrower job.

### context rot  [KON-tekst rot]  noun
The decay of earlier decisions in a long session, as they fall out of the model's effective attention.
Also called context decay. The tell is an agent that silently contradicts a constraint you set forty turns ago, with no sign it ever knew.

### context window  [KON-tekst WIN-doh]  noun
The total number of tokens a model can attend to at once.
Bigger is not free. Cost, latency, and the difficulty of finding the relevant part all rise with it, which is why the skill is selection rather than accumulation.

### demo cliff  [DEM-oh klif]  noun
The point at which a prototype that demonstrated beautifully cannot be extended into a product.
Everything before the cliff was generated; nothing before the cliff was designed. The fall is usually discovered when the second feature has to share state with the first.

### EARS notation  [eerz noh-TEY-shuhn]  noun
Easy Approach to Requirements Syntax: five sentence patterns that turn vague requirements into testable, machine-parseable statements.
Created by Alistair Mavin and colleagues at Rolls-Royce and first published at the IEEE requirements engineering conference in 2009, long before anyone was prompting anything. The patterns are ubiquitous, event-driven, state-driven, optional-feature, and unwanted-behaviour. It is the cheapest upgrade available to most vibe coaders.

### embedding  [em-BED-ing]  noun
A numeric vector representing the meaning of a piece of text, such that similar meanings sit near each other.
The machinery under semantic search. Similar is not the same as relevant, which is why a second ranking pass usually earns its keep.

### eval-driven development  [ih-VAL-driv-uhn dih-VEL-uhp-muhnt]  noun
Writing the evaluation before the feature, so you can tell whether a change improved anything.
Test-driven development adapted for systems that do not produce the same answer twice. Without it, "the new prompt seems better" is the only quality signal you have, and it is not one.

### evals  [ee-VALZ]  noun
Repeatable tests that score model or agent output against expected behavior.
Your regression suite for a system that will not give the same answer twice. The maturity move is replacing "that looked right" with a number you can watch over time.

### fine-tuning  [fahyn-TOO-ning]  noun
Further training of an existing model on specific examples to specialize its behavior.
Usually the wrong first move. Most problems people bring to fine-tuning turn out to be retrieval or prompting problems that were cheaper to fix.

### greenfield  [GREEN-feeld]  noun
A project with no existing code, where the agent can generate freely.
Where the demos come from. Greenfield is largely solved, which is why greenfield demos impress nobody who has shipped anything.

### hallucination  [huh-loo-suh-NEY-shuhn]  noun
A confident fabrication: a function, citation, parameter, or fact that does not exist.
The word is doing PR work for the machine, implying a perceptual error rather than a generative one. The model is not seeing things; it is producing plausible text, and sometimes plausible text is false.

### harness  [HAR-nis]  noun
Everything wrapped around a base model to make it an agent: system prompts, tool wiring, control flow, retries, and memory glue.
Also called scaffolding. Increasingly, gains on long-horizon tasks come from a better harness rather than a smarter model, which is why practitioners talk about the harness more than they talk about the weights.

### human verification gate  [HYOO-muhn ver-uh-fi-KEY-shuhn geyt]  noun
The named point in a pipeline at which a person is accountable for what ships.
If you cannot point to yours, you do not have one. The gate is a person and a moment, not a policy document.

### intent drift  [in-TENT drift]  noun
The gap between what you asked for and what the model reasonably assumed.
"Add login" is wildly underspecified, and the defaults a model picks to fill the gap are rarely yours. Drift is not the model being wrong; it is the model being reasonable about something you never said.

### the loop  [thuh loop]  noun
The core cycle of vibe coding: generate, evaluate, refine, repeat, until output converges on intent.
The whole practice lives here. Everyone shows you pass one; almost nobody publishes pass four, which is where you find out whether you understood the problem.

### MCP  [em-see-PEE]  noun
Model Context Protocol: an open standard for connecting AI agents to external tools and data sources.
Introduced by Anthropic in late 2024 and now the de facto plug standard. If a tool does not speak it, that tool is effectively outside the ecosystem.

### mise en place  [MEEZ ahn PLAHS]  noun
Assembling all context, constraints, and reference material before the first prompt.
Borrowed from professional kitchens, where it means everything is on the board before the burner goes on. The failure mode it prevents is identical in both trades: improvising mid-service because you did not prep.

### model routing  [MOD-l ROO-ting]  noun
Sending different work to different models: cheap and fast for boilerplate, frontier for architecture, specialized for security review.
The tech-lead move applied to machines. Once you route, you stop having a favorite model and start having a workflow, and the orchestration layer quietly becomes your real IDE.

### one-person product team  [WUHN-per-suhn PROD-uhkt teem]  noun
A single builder covering product, architecture, implementation, deployment, and support at a scope that recently required a small team.
The defining archetype of the era, and usually a rotation through several working styles rather than a single one. The bar for shippable rose at the same time the headcount fell.

### overfitting to the prompt  [OH-ver-fit-ing]  noun
Producing output that satisfies the literal request while missing its purpose.
You asked for the button to be centered and now nothing else on the page is. Technically compliant, practically useless.

### prompt caching  [prompt KASH-ing]  noun
Reusing a fixed prefix across many model calls to cut latency and cost.
The single biggest cost lever most solo builders never switch on. Put the stable material first and the variable material last, and the savings arrive without changing a word of the prompt.

### prompt injection  [prompt in-JEK-shuhn]  noun
An attack in which untrusted content carries instructions that an agent then follows.
The content can be a web page, a file, a code comment, or a support ticket. Harmless when the agent can only talk; genuinely dangerous the moment it can act.

### prompt jockey  [prompt JOK-ee]  noun
Mildly derisive term for someone who prompts constantly and verifies nothing.
The insult is not about skill level. It is about the missing second half of the job.

### RAG  [rag]  noun
Retrieval-augmented generation: pulling relevant documents into the context window at query time instead of relying on trained-in knowledge.
Introduced in a 2020 research paper and now the standard way to make a general model answer questions about your particular data. Most RAG failures are retrieval failures wearing a generation costume.

### Ralph loop  [ralf loop]  noun
Running an agent repeatedly on the same task until the output converges.
Named for a pattern popularized by Geoffrey Huntley. It works better than it should, and it works best when the task has a checkable finish condition; without one, it loops forever, confidently.

### refinement pass  [ri-FAHYN-muhnt pas]  noun
One iteration of feeding a model its own output plus a correction.
Passes are cheap and non-linear. Passes one and two usually fix cosmetics; the structural error tends to surface on three or four, once the obvious problems stop hiding it.

### reranking  [ree-RANG-king]  noun
A second scoring pass that reorders retrieved candidates by relevance to the actual question.
Retrieve broadly, rerank narrowly. Top-five by raw vector similarity is frequently worse than top-twenty reranked down to five.

### RPI  [ahr-pee-AHY]  noun
Research, Plan, Implement: a three-phase agentic method in which the agent studies the codebase and writes a plan before it is allowed to edit anything.
Attributed to Dex Horthy. The discipline is in refusing to let the agent start typing during the research phase, which it will always want to do.

### rubber duck debugging  [RUHB-er duhk dee-BUHG-ing]  noun
Explaining a problem aloud, line by line, to an inanimate object until the flaw becomes obvious.
Popularized by The Pragmatic Programmer in 1999. The duck has been promoted: it now answers, which is an improvement roughly half the time.

### silent regression  [SAHY-luhnt ri-GRESH-uhn]  noun
A change that fixes the reported bug and quietly breaks a working behavior nearby.
The reason later refinement passes need the same test discipline as the first. The model optimizes what you pointed at and has no stake in what you did not.

### slop PR  [slop pee-AHR]  noun
An unrequested, agent-authored pull request that costs a maintainer more to review than the fix would have cost to write.
The externality of cheap generation. The cost does not disappear; it moves to whoever has to read it.

### slopsquatting  [SLOP-skwot-ing]  noun
The practice of registering a package name that AI models commonly hallucinate, so that the fabricated import resolves to attacker-controlled code.
Built from typosquatting, where the bait is a misspelling; here the bait is the model's imagination. It is the reason to verify every unfamiliar import against your installed version before reading a single line of the implementation.

### spec-code drift  [spek-kohd drift]  noun
The divergence that accumulates when code evolves and its specification does not.
Invisible until it is expensive. The spec stops describing the system, and everyone keeps consulting it anyway.

### spec-driven development  [spek-DRIV-uhn dih-VEL-uhp-muhnt]  noun
A method in which a versioned, structured specification is the source of truth and code is a generated artifact verified against it.
Usually shortened to SDD and run in four phases: specify, plan, tasks, implement, with a human checkpoint at each. It went mainstream because agents write code well and guess intent badly.

### subagent  [SUHB-ey-juhnt]  noun
A scoped agent spawned for one narrow task, running its own context and reporting a result back.
The main structural defense against context explosion. The parent keeps the plan; the child keeps the mess.

### success criteria  [suhk-SES krahy-TEER-ee-uh]  noun
The observable, checkable conditions that define a finished piece of work.
Distinct from a description of the feature. "A user from tenant A receives zero rows from tenant B, proven by a test" is a success criterion. "Make it secure" is a wish.

### system prompt  [SIS-tuhm prompt]  noun
Standing instructions that apply to every turn of a conversation, separate from the user's message.
Where role, constraints, and refusals live. Everything in it is paid for on every single call, which is an argument for editing it like a lease.

### temperature  [TEM-per-uh-cher]  noun
A setting controlling how much randomness a model applies when choosing its next token.
Low temperature for code and extraction, higher for ideas. It does not make a model more or less correct, only more or less predictable.

### token  [TOH-kuhn]  noun
The unit a model reads and writes: roughly a short word or word fragment.
Everything is priced, limited, and measured in tokens, which is why the practical unit of an AI system is not the sentence but the fragment.

### tool use  [tool yooz]  noun
A model invoking defined external functions rather than only producing text.
Also called function calling. It is the line between a model that describes an action and an agent that takes one, and it is the line at which prompt injection stops being theoretical.

### the 90% trap  [thuh NAHYN-tee per-SENT trap]  noun
The pattern in which an agent produces ninety percent of a feature in an hour and the last ten percent takes days.
The remaining work is hard precisely because you did not build the first ninety percent and therefore cannot navigate it. The time saved was borrowed, not earned.

### trust paradox  [truhst PAR-uh-doks]  noun
The condition of near-universal daily use of AI coding tools alongside falling trust in their output.
In Stack Overflow's 2025 survey, adoption climbed to eighty-four percent while active distrust of accuracy rose from thirty-one to forty-six percent in a single year. The gap between what people use and what people believe is the whole opportunity.

### verify ruthlessly  [VER-uh-fahy ROOTH-lis-lee]  phrase
The community's nearest thing to a motto: use the tools aggressively, then check the output without mercy.
The compressed version is shorter still. It runs is not it works.

### vibe coader  [vahyb KOH-der]  noun
A builder who produces working software primarily by directing AI agents rather than typing implementation code, and who accepts full ownership of the result anyway.
The extra letter is load-bearing. A coder writes the code; a coader conducts it. The craft is not syntax but framing, constraint, and verification, and the measure of one is not how much code they generate but how little they have to throw away.

### vibe coding  [vahyb KOH-ding]  noun
Building software by describing intent to an AI model and accepting the generated code with little or no line-by-line review.
Coined by Andrej Karpathy in February 2025, in the spirit of giving in to the vibes and forgetting the code exists. It was a description of low-stakes weekend hacking. It has since been stretched to cover serious production work, which is why roughly half of every argument about vibe coding is actually a disagreement about the definition. Collins named it their 2025 Word of the Year; the same year, Merriam-Webster picked slop.

### vibe debt  [vahyb det]  noun
Technical debt accumulated at generation speed.
Ships in an afternoon and compounds for a year. Ordinary technical debt is incurred by a person who knew they were cutting a corner; vibe debt is incurred by nobody in particular.

### vibe engineering  [vahyb en-juh-NEER-ing]  noun
Proposed successor term for the disciplined practice: the same tools, with actual review.
Coined largely by people tired of explaining that they do read the code. Expect the casual sense of vibe coding to survive and the professional sense to keep shopping for a new name.

### went badly  [went BAD-lee]  phrase
Category label for a build, prompt, or session that failed, published with the diagnosis intact.
The most credible thing a practitioner can publish and the rarest. Everyone shows the finished build; the failure is where the transferable knowledge actually lives.

### yak shaving  [yak SHEY-ving]  noun
A chain of prerequisite tasks that carries you further and further from the thing you set out to do.
The term traces to MIT AI Lab lore of the 1990s. Agents are unusually good at yak shaving, because each detour looks locally reasonable and nothing in the loop asks whether you are still solving the original problem.

## The Toolbox

### Learn
- Claude Code Workflows (docs.claude.com) — Recipes for debugging, refactoring, and PRs with Claude Code
- CodeQL (codeql.github.com) — Scans your code for vulnerabilities automatically
- Codex Docs (learn.chatgpt.com) — Planning, prompting, testing, and permissions for Codex
- CS50x (cs50.harvard.edu) — Harvard's intro course. The deep one, when you're ready
- Cursor Docs (docs.cursor.com) — How to set up rules that stick across every Cursor session
- Dependabot (docs.github.com) — Auto-patches dependencies with known security holes
- FastAPI Tutorial (fastapi.tiangolo.com) — Routes, validation, and auth for a Python backend
- GitHub Skills (skills.github.com) — Learn Git and GitHub by doing, not by reading
- MDN Learn (developer.mozilla.org) — The reference for how the web actually works
- Next.js Learn (nextjs.org) — Build a real app while learning routing and server components
- OWASP ASVS (owasp.org) — The security checklist to hold your app against
- Playwright (playwright.dev) — How to write browser tests that do not flake
- PostgreSQL Tutorial (postgresql.org) — Tables, joins, and indexes from the source
- Pro Git (git-scm.com) — The branching chapter. Read that one, skip the rest
- React Foundations (react.dev) — State, props, and rendering, explained properly
- Sentry Docs (docs.sentry.io) — See the errors your users hit before they email you
- v0 (v0.app) — Describe a UI, get a working React screen you can keep

### Agents & IDEs
- Arena (arena.ai) — Benchmark models head to head
- Hugging Face (huggingface.co) — Where the open models, datasets, and demos live
- DeepSite (huggingface.co) — Describe a site, get a working single-page build you can download — a Hugging Face Space
- Latent Box (latentbox.com) — A big curated index of free AI tools
- LMArena (lmarena.ai) — Blind model battles. The honest answer to "which is better"
- Manus (manus.im) — An agent that plans and does multi-step work in a browser
- PentestGPT (pentestgpt.ai) — AI-assisted security testing on your own stack
- Poe (poe.com) — Many models behind one chat window
- Readdy (readdy.ai) — Prompt to website, good for a first pass
- Rocket (rocket.new) — Describe an app in plain text, get a full-stack build
- Runable (runable.com) — Chain AI steps into a workflow you can rerun

### Generate
- Carto Art (carto-art.vercel.app) — Turn any map into a printable poster
- ContentCore (contentcore.xyz) — 3D content and design assets
- Google Flow (labs.google) — Google's Veo video generation lab
- Higgsfield (higgsfield.ai) — Make short cinematic video from a prompt
- Wan 2.2 Animate (huggingface.co) — Animate a still character from a reference video — open-weights demo on Hugging Face
- AI QR Code Generator (huggingface.co) — QR codes that are also pictures and still scan — a Hugging Face Space
- Inkarnate (inkarnate.com) — Draw fantasy maps in the browser
- Motion BGs (motionbgs.com) — Free 4K animated backgrounds
- Perchance (perchance.org) — Free text-to-image, no signup
- Recraft (recraft.ai) — Image generation built for designers, with vector output
- Split My Song (splitmysong.com) — Split any song into separate instrument tracks
- Zoo CAD (zoo.dev) — Type a description, get a real CAD model

### Components
- Color Hunt (colorhunt.co) — Curated palettes, faster than arguing with a model
- Faces (faces.app) — Build a presentation or site by dragging things around
- Lordicon (lordicon.com) — Animated icons without writing a single keyframe
- Pictogram2 (pictogram2.com) — Free stick-figure pictograms for diagrams
- PNGWing (pngwing.com) — Transparent PNGs for whatever you are mocking up
- React Bits (reactbits.dev) — Animated React components you can paste in today
- Schultz Schultz (schultzschultz.com) — Odd, lovely browser design tools from a real studio
- Space Type Generator (spacetypegenerator.com) — Play with type layouts in the browser
- Tooooools (tooooools.app) — Turn any image into stipple, dither, or ASCII
- TopoExport (topoexport.com) — Download vector maps and 3D terrain models
- UIBall (uiball.com) — Loaders and spinners a generated build always forgets

### Research
- Apify (apify.com) — Managed scraping so you skip writing the browser harness
- Car-Part (car-part.com) — 200M+ used auto parts, searchable
- Class Central (classcentral.com) — Find free university courses in one place
- GPT Excel (gptexcel.uk) — Plain English into a working spreadsheet formula
- iFixit (ifixit.com) — Free repair guides for almost any device
- Learn Anything (learn-anything.xyz) — A visual map of how topics connect
- Live Threat Map (threatmap.checkpoint.com) — Live map of cyber attacks happening now
- Magazine Rack (archive.org) — Decades of digitized magazines, free
- Napkin (napkin.ai) — Paste text, get a diagram
- NoBook Chem (chemistry-en.nobook.com) — Run chemistry experiments in a virtual lab
- PhET Sims (phet.colorado.edu) — Interactive physics and chemistry simulations
- PlayPhrase (playphrase.me) — Search movie clips by the line of dialogue
- Shodan (shodan.io) — Search engine for exposed devices. Check your own
- Visualping (visualping.io) — Watches a page, pings you when it changes

### Utilities
- 3DTuning (3dtuning.com) — Configure and customize cars in 3D
- DOF Simulator (dofsimulator.net) — See how aperture changes depth of field
- Dupe (dupe.com) — Find the cheaper version of an expensive thing
- Enter.pro (enter.pro) — Workspace and project tool
- MagicV (magicv.art) — AI resume builder and optimizer
- Make My Drive Fun (makemydrivefun.com) — Finds interesting stops along your route
- McStumble (mcstumble.com) — Text and productivity tools, no signup
- Fake Text Me (mcstumble.com) — Mock up an iMessage screenshot for UI demos and tests
- Mr Free Tools (mrfreetools.com) — A pile of small free utilities
- MyFridgeFood (myfridgefood.com) — Tell it what you have, it tells you what to cook
- ShortURL (shorturl.at) — Shorten a link, nothing else
- White Screen (whitescreen.online) — A blank white screen. Surprisingly useful

### Rabbit Holes
- Atom Animation (atomanimation.com) — Watch atoms and molecules move
- August (meetaugust.ai) — AI health assistant
- Draw a Stickman (drawastickman.com) — Draw a stickman and send him on an adventure
- Elastic Man (adultswim.com) — Stretch a face around. You will lose ten minutes
- Emupedia (emupedia.net) — Classic games and old operating systems, emulated
- Floor 796 (floor796.com) — An endless animated pixel scene full of references
- Form Follows Function (fff.cmiscm.com) — An interactive art piece about form and function
- GeoFS (geo-fs.com) — A real flight simulator in your browser
- Italian Hall (italianhall.org) — A museum exhibit, online
- Meta SAM 3D (ai.meta.com) — Meta's 3D reconstruction research
- Muted (muted.io) — Learn music theory by clicking things
- My Retro TVs (myretrotvs.com) — Watch things as if on a vintage TV set
- Neal.fun (neal.fun) — Scroll through the scale of life
- Rent a Human (rentahuman.ai) — Hire a human through an AI interface
- Rotating Sandwiches (rotatingsandwiches.com) — Sandwiches. Rotating. That is the whole site
- Slow Roads (slowroads.io) — Drive an endless road. Nothing happens. It rules
- The Toymaker (thetoymaker.com) — Printable paper toys you cut out and fold
- Toy Theater (toytheater.com) — Simple educational games for kids