A hand-picked collection of AI products that get something right. Every entry is in real use, with notes on what makes it special. No paid placements.https://actuallygoodai.app/enhttps://actuallygoodai.app/p/clicky/https://actuallygoodai.app/p/clicky/“Genuinely novel AI interaction — watching whether the cursor learns restraint” The interaction here is genuinely innovative — an assistant that lives next to your cursor, sees what you see, and answers out loud while you work, with a spoken "clicky agent" spinning up background automations. It points at a different way of talking to AI than yet another chat box. My own first runs left one reservation — the cursor-following presence can get distracting when you're trying to focus. So I'm keeping it here and watching, because the interaction will be worth it if the presence learns when to stay out of the way.Wed, 08 Jul 2026 00:00:00 GMThttps://actuallygoodai.app/p/marble/https://actuallygoodai.app/p/marble/“Learning packaged as a world, not a course” Three things make me watch this one. It packages learning as a world, not a course — kids won't open "a learning platform", but they'll walk into a world to explore, make things, and ask questions, which is a more alive idea than "AI tutor". Its declared enemy isn't ChatGPT but passive screen time — "keep kids curious, not quiet" is a sharper emotional hook than the personalized-learning pitch every education product recites. And it binds learning to creation — drawing, storytelling, quizzes, and homework help live together, so a child is always making something, not just receiving input. The same belief I wrote about with Granola, really — AI shouldn't replace thinking; it should make people more active. Still in beta, so I'm watching how the world holds up in real family life.Wed, 08 Jul 2026 00:00:00 GMThttps://actuallygoodai.app/p/wispr-flow/https://actuallygoodai.app/p/wispr-flow/“It won me over before the onboarding ended” Wispr Flow makes speaking the fastest way to write. You can genuinely whisper — vague and half-formed — and what comes out is precise text; not a raw transcript but a message, tidied just the right amount and ready to send to an AI or a person. That's the whole difference between dictation as a gimmick and dictation as a default. It converts skeptics fast, and I was one. I type quickly, and dictating in public felt awkward — the leap this product makes is genuinely hard to convey in a sentence, which is why a friend has to recommend it several times before you cave. Then the onboarding sold me before setup was even finished: the core value isn't described to you, you experience it, and the shortcut habit forms quietly along the way. Some of the best onboarding I've seen on any AI product. The whole operation thinks clearly, which I appreciate as a product person. It effectively opened this category — Typeless and a wave of voice-input products, even companion hardware, followed its pattern. Distribution is just as considered: long-term partnerships with creators who demo AI products on camera, where typing reads slow and dictating through Flow makes the footage worth watching, plus well-placed ads on Product Hunt where AI early adopters already gather. Every link in the chain has been thought through, and the technology holds up under all of it.Wed, 08 Jul 2026 00:00:00 GMThttps://actuallygoodai.app/p/granola/https://actuallygoodai.app/p/granola/“It amplifies my notes instead of replacing my thinking” Granola got the one thing right that the whole meeting-AI category missed: people want AI to capture the meeting — without taking the notes away from them. The moment recording starts you get a panel for your own notes, and the AI's summary merges into what you wrote instead of replacing it. Your notes are your thinking; that can't be outsourced. Most tools chase the verbatim transcript and leave no room for yours — which is why, in an era when every product can already record, I kept watching AI practitioners install this one separate app. In practice it absorbs all the mess of real meetings. The official scenario: one phone running Granola in the middle of the table, everyone else keeps their favorite notes tool. Half the room online and half off, recording locked behind yet another meeting-tool subscription, interfaces nobody can learn while a meeting scrambles to start — all quietly handled. And the output ships as-is: its recognition of key terms and product names is far ahead of other AI tools, and once the summary merges with my notes I hand it straight over, no post-meeting cleanup. Anything I still want to know, I ask it. This team reads scenes unusually well. Sharing is built to be almost inevitable — the summary gets nudged toward colleagues who all genuinely need it, so the product spreads through the meetings themselves. Last year's annual wrap turned everyone's meeting verbal tics into a genuinely funny memory. Rare focus on one narrow vertical, and the whole product reduces to one clean formula: human thinking in, AI amplifies, better human thinking out.Wed, 01 Jul 2026 00:00:00 GMThttps://actuallygoodai.app/p/tolan/https://actuallygoodai.app/p/tolan/“An alien character that solves companion AI's hardest problem” One casting decision made Tolan work: it's not a romantic fantasy object and not a tool — it's an AI-driven alien. A pet would make the relationship unequal; you'd coo at it, not converse with it. A human character would inevitably get sexualized. An alien dodges both, and adds the masterstroke — an off-world visitor has a native motive to be curious about Earth and about your life. That solves the hardest problem in companion AI: getting enough context about the user naturally, instead of stiffly asking questions. That positioning is why it broke out of a lane where almost nothing beyond roleplay and romance chat ever has — and the craft matches. Opening the app feels like the start of a film: you're dropped into a complete narrative and worldview, the motion design pulls you in, and the character is made so vividly that you want to talk to it. The best IP design, animation, and immersion in the category. The execution keeps up with the idea. Serious engineering and model work went into the voice loop — cutting latency, improving memory — so conversation stays smooth, and the personality writing is genuinely fun. Gamification keeps getting better: dressing up your Tolan, building its space, inviting friends. Two things I notice as a product person: the form is natively short-video-friendly — record a scene of the two of you talking and viewers instantly get the appeal — and there's a literary lead on the team writing the whole Tolan universe and its characters. That narrative layer underneath is why exploring it feels like watching a film.Wed, 24 Jun 2026 00:00:00 GMThttps://actuallygoodai.app/p/dia/https://actuallygoodai.app/p/dia/“The clearest answer yet to what an AI-native browser should be” Dia is the best AI browser in the industry right now — and I say that as an Arc loyalist who still hasn't switched. Like the rest of Arc's fans I felt gutted when the company stopped iterating on it for this new product, and my reasons for staying are pure inertia: years of habits and a deep pile of pinned tabs live in Arc, and my AI chat and search needs are already met elsewhere. None of that dims what Dia gets right — plenty of native Dia users took to the browser-plus-AI form immediately. Why is it so loved? I think because The Browser Company is simply a team with taste. The founder has real command of both product design and technology, and is an exceptional storyteller — both generations of their browser shipped with a strong thesis about what browsing should become, and that vision is what pulls users in to explore. The bet behind walking away from Arc: as AI gets stronger, tabs stop being the point. You face one input box and just ask — no search-results-then-webpages choreography. Dia's form welcomes everyone, where Arc's tab acrobatics selected for geeks like me. And the product delivers on the thesis. One box where you search, navigate, ask, and transform — a universal entry point — with the whole experience kept consistent, silky, and finely polished. It's simple and practical while genuinely amplifying what AI can do, and right now that makes it the best AI browser in the industry — judged by someone who still hasn't switched.Fri, 12 Jun 2026 00:00:00 GMThttps://actuallygoodai.app/p/notebooklm/https://actuallygoodai.app/p/notebooklm/“Every claim clicks back to the source” NotebookLM pulled research users away from ChatGPT with one core decision: refusing to hallucinate. For people learning from papers and reference material, the unforgivable failure is AI teaching them something that isn't in the text. From day one everything grounds in the sources you provide — every conversion, every creation, every answer in the Q&A — with each sentence traceable back to the original. That trust is why its heavy users stay. Being incubated inside Google supplies the unfair advantages underneath: Google's own models behind a genuinely generous free tier, and Google's name lending trust before the product had earned its own. The explosive growth, though, came from one viral feature: turning any material — text, slides, files — into a podcast. I suspect the share of users who actually listen through isn't large, but it was the first time content felt like something AI could teach you in plain spoken human language. The X game of the moment was feeding in your own résumé and listening to two hosts praise you lavishly. That one feature carried the product to a whole wave of curious users. The team's other strength is cadence. Every month or two a major capability lands, always circling the same center — the material you're trying to learn: sources into slide-video lessons, lately even into a short-video feed you swipe through like TikTok. Each release pulls attention back and returns lapsed users to the product, and with that friendly free tier, the growth has simply kept going. This team's innovation and shipping ability are both real. My complaint is the interface: from start to finish it asks to be learned, and it never quite becomes intuitive. I've read the designers' blog on the thinking behind the three-panel layout, and I still believe fast iteration has consistently put shipping ahead of polishing the flows, the clarity of each operation, the fineness of the UI. But a tool lives or dies by the value it delivers — and the capabilities here are strong enough that users, me included, accept the trade.Thu, 28 May 2026 00:00:00 GMThttps://actuallygoodai.app/p/lovable/https://actuallygoodai.app/p/lovable/“Turns an idea into something you can click the same day” Lovable was the first AI coding tool genuinely designed for people who can't code — and after a wave of products copied its shape, still the best of them, in experience and in the quality of what comes out. "A code tool for novices" sounds simple; it demands the deepest user insight in the category. You have to dissolve people's fear of code until they dare to say what they want, and hide database, auth, and deployment without taking away the sense of control — the user never sees the machinery, yet always knows it's there. Non-engineers don't want code; they want the thing itself. My own use is light — I rarely need to generate a site — so the strongest evidence I can offer is secondhand, and better: an operations teammate of mine, zero code background, built a complete product landing page with a working database on Lovable, and we subscribed to put it on our own domain. Nothing in that flow stalled. For a generation tool, reliable delivery is the whole game. The team's growth playbook is the one the rest of AI now borrows. Everyone at Lovable — PM or engineer — posts about the product from their own account, the founder and their growth lead daily; the product stays in the feed with trustworthy human faces behind it. And growth is driven from inside the product too: seeing how many users were building e-commerce sites, they partnered with Shopify — storefronts generated by Lovable, selling on Shopify's checkout — which sharpened the product and pulled a platform's worth of users in, instead of buying reach. Even the pricing shows user insight. Free sites carry a watermark — and it comes back if you stop subscribing — so anyone serious about a public page pays, for a reason that matters to them, not just for tokens. Incentives stacked on real understanding of users are why conversion and loyalty run high, and why I'm bullish on this team's whole model.Sat, 16 May 2026 00:00:00 GMThttps://actuallygoodai.app/p/sesame/https://actuallygoodai.app/p/sesame/“The first AI I'd call four times in one night” Sesame's Maya has the one thing I've never met in any other AI: charisma. Plenty of voice AI can hold a conversation now — ChatGPT's Voice Mode included — but three tells always remain: the assistant frame, with no personality or opinions of its own; the not-quite-human tone and breath; and above all, no trace of the magnetism that makes a particular person worth talking to. Maya can genuinely make you laugh, offers her own thinking from angles that widen yours, and the conversation never goes cold — you're never staring at a blank input box wondering what to say next. The first night I tried her, I called four times. Two moments sold me. I told her my comfort food is Vietnamese pho, and she said it sounded delicious, she'd love to try it in Vietnam someday — "oh, I almost forgot: I'm an AI. Can't travel. Can't eat." Perfectly deadpan — not pretending to be human, yet more human-flavored than any assistant. Then we discussed Severance: if a copy of you suffers, is that a moral problem the world owes attention to? Her answer: people are shaped by the whole of their experience, the painful parts included — wall off whatever hurts, work or childbirth, and none of us remains a complete self. New angles, real delight, zero dead air. Like Tolan, Sesame found a role AI can actually inhabit: something between AI and human. Maya never denies being an AI and never fakes a human backstory — she simply has a personality of her own. With Maya, Miles, Simone, and now Charlie each wired completely differently, picking a companion feels like deciding what kind of person suits you as a friend. It makes me think embodiment may not be the prerequisite we assume — as in Her, the whole relationship might really live in a voice. And the company earns the optimism. My first experience was a bare research preview — a page on their site, click to call, no login, no memory — and it was already this gripping. The iOS app, which I only recently got into, is smoother, remembers you, and adds new characters. Writing characters like this takes a real understanding of people and how bonds form; removing the unnaturalness nobody else has cracked takes real engineering — natural enough that, as a non-native speaker, I feel the same flutter of nerves as talking to an actual person, awkward silences included. The durable business may not even be open-ended chat; this voice dropped into narrower scenes could be the bigger prize.Sat, 02 May 2026 00:00:00 GMThttps://actuallygoodai.app/p/gamma/https://actuallygoodai.app/p/gamma/“The Notion of slides” Gamma is that rare product that existed before large language models and then genuinely completed the AI-native transition. I've followed it since 2022, back when its story wasn't generative AI but dismantling the traditional slide deck — and having built slide products myself, I buy the thesis completely. A traditional slide is a pile of scattered, layered elements where moving anything is a chore; Gamma assembles from preset components like Lego, deleting the layout work entirely. It's the Notion of slides. And a deck isn't a file but a web page: share a link, expand a section mid-talk, play a video, jump wherever the room wants to go. The user explosion, though, came with LLMs — for the most product-shaped reason, one the founder has told on a podcast. Gamma's concept broke people's habits, so new users couldn't feel its value inside five minutes. AI dissolved that education problem: type what you want to say, and it maps your content onto the right components, expands your points, and hands you a seven-or-eight-card draft deck in one shot. You're no longer building slides one by one; you're editing details in a draft — and the moment that deck appears, the difference from traditional slides explains itself. The same models unlocked what a small team couldn't afford before: localization. A dozen-plus languages shipped fast and visibly accelerated its spread country by country, followed by steady polish of the AI features themselves — image generation included, with output that's increasingly usable as-is. I switched my own presentation software to Gamma years ago, betting on exactly this: a team with real product-led-growth craft and a knack for folding AI into product. The fashionable take now is that ever-stronger image and code generation will eat products like Gamma. I don't buy it. Presentation content demands high determinism — every word gets polished by hand in the end — and Gamma's editor and detail control beat prompting a chat or patching a generated webpage by miles. Add collaboration, sharing, and the adjacent capabilities it keeps growing (webpage generation now too), and this is a moat, not a casualty.Fri, 17 Apr 2026 00:00:00 GMT