Connect with the Product Hunt community

I tried spinning up multiple AI coding tabs last weekend to knock out a bunch of bug fixes in one go thinking it would make me four times as productive. Instead I spent two hours staring at different git diffs completely losing track of what logic belonged where.

The agents spat out code way faster than I could read, let alone test or understand. Code generation speed doesn't mean much if your own review bottleneck slows everything down to a crawl.

How many parallel tasks do you actually let your tools run before you start missing bugs?

Thank you, community members for making us #1 in the PH launch in July 2026.

Our new version is currently live on product hunt. Please take a look and support us with your upvotes, comments and please also try out the product.

Hi Hunters,

Vladimir here, co-founder of Intrascope. Today is our third launch on Product Hunt, and 2.0 is the biggest update we have shipped so far.

The problem we keep hearing from companies: every team uses a different AI tool, nobody knows what it costs, and company context gets pasted into chats again and again.

Intrascope gives the whole team one workspace with 120+ models (OpenAI, Anthropic, Google, Mistral, xAI and more), while admins keep control over access, budgets and usage.

Claude renamed its plans on October 30. Max 20x is now Ultra and Max 5x is now Max. Ultra has no 5-hour limit any more, only the weekly one, so the ring on my bar went empty. In 0.3.33 the ring follows your weekly limit when there's no 5-hour one, and Plan limits shows just the weekly bar.

We launch a feature here on Thursday, and the landing page section for it is the one new users keep bringing up on calls and in replies, more than the hero, more than pricing.

It is a simple block. One sentence: see the change before it ships. One picture of a working copy of your site with the change on it, and two buttons: approve, or send it back.

I think it works because it answers the question people actually have about any tool that writes code for them: what if it gets it wrong? The answer is not trust us , it is look at it first . The other sections explain what it does. This one removes the fear.

Curious what the equivalent section is on your landing page: the one that answers the fear rather than describing the feature. And for those who have run the test, did it convert better than the feature description?

When you call customers, the first thing you often hear is a call screener, voicemail, an automated menu, or a "number switched off" message. In our own test of about 2,700 real calls, that was nearly 18%.

An agent who doesn't notice starts talking to a machine, wasting money and sounding silly.

Now it recognises what it is talking to and reacts:

- Call screener: introduces itself, then waits for the person.

We've been working on Dots, a macOS app that turns your Mac's notch into a customizable workspace.

One of the features we've added is drag and drop.

The idea is pretty simple. Instead of opening another app or switching windows, you can just drag files into your notch and access tools like file conversion right there.

We built Dots natively using Swift and SwiftUI because we wanted it to feel like a natural extension of macOS rather than another window sitting on your desktop.

We re giving AI tools and other tech companies more and more data.
And it s not just our names and email addresses.

Sometimes, it gets a little too personal:

Phone numbers. I hate getting SMS messages, especially since I only use my phone when absolutely necessary. (Family calls and important things related to doctors)

Biometric data. Like fingerprints.

Our callers mix Hindi and English in the same breath: "haan toh mein bol raha tha, why are you calling me?" We assumed language handling was a detection problem. It turned out to be about six separate problems, and we found them one at a time, in production.

1. Detecting a language switch without flip-flopping
A caller says one English word and the bot jumps languages. We settled on a heuristic switch with guardrails you can configure per agent:

• a minimum number of words per turn (default 3)

• the same language for consecutive turns (default 2)

• a dominant-language threshold (0.6) from a fastText classifier

• a reset of the counter on mismatch, so one stray turn doesn't count

• "meta" references ignored, for example someone saying "English" while speaking Hindi

2. Callers who just tell you what they want
"Let's speak in English" shouldn't need two confirming turns. We added explicit request detection, in English and Hindi (e.g. " "). It bypasses the heuristics and switches immediately.

We already are getting a overwhelming response in ourr website and product and people are loving it, try it out first hand
https://platform.vaani.dev/
Grab your free credits, and this thread and our socials are open for feedback and comments, Thank you so much!

A voice agent that makes up answers is worst feeling you can give to the callers. Callers ask about prices, policies and product details, and a wrong answer costs trust. With our Knowledge Base feature, you upload your files, paste text or point it at your website. During a call, the agent looks up the answer in your material and uses it. This works with long, multi-page documents. Lets you to add or update content without rewriting the agent and you can switch it off for agents that do not need lookups.

A short pause can happen while it looks something up. The answer is worth the wait.

What is the one document you would upload first?

Building a team or want to join a startup?

Founders, teams, and startups drop a comment if you're hiring.

I was 15 when I entered university...

In my 4th year of engineering, I dropped out to go all-in on Murmell.

AI agents are getting pretty good at finding information, summarizing documents, and answering questions. But what about letting them change the information itself?

Imagine an AI agent finds outdated onboarding instructions in your company's wiki. Should it be allowed to fix them immediately? Suggest an edit for someone to review? Or simply flag the problem and leave everything to a human?

We've been thinking about this while developing the XWiki MCP Server.

Our approach is to respect existing user permissions and keep write operations disabled by default. Administrators can enable them when needed.

Here's how webhooks work in Vaani (no code needed)

A webhook is just Vaani sending your app a message when something happens on a call.

You pick the events you care about: call_started, call_ringing, user_picked_up_at, call_rejected, call_no_answer, call_failed, call_ended.

You give us a URL. When a call ends, we send it a message, and you can drop the result in a Google Sheet, ping Slack or update your CRM.

I feel like most people on Product Hunt are too afraid to run paid ads.
Meanwhile, it's been one of the biggest drivers of growth at all the startups and projects I've worked on.
I'm super curious on what everyone's relationship is like with paid ads?
We've rebuilt Fastlane from the ground up to be the world's first end to end AI agent for paid ads, so would love to understand this especially!

Lancetta 0.64.0 is out: Lancetta is on your desktop now.

A small widget for Claude and one for Codex, with both windows and the time to each reset, and a medium one with every agent and the suggestion that matters most.

Early on, I measured success by daily signups. But once we monetized, watching paid users cancel hit totally differently.

The hardest part was realizing that adding quick features to our backlog won't fix most of these cancellations. People aren't leaving because a feature is missing they're leaving because the product isn't smoothly inti their existing habits.

For founders managing paid tiers: how do you adjust your roadmap to fix fundamental workflow friction?

Over the last year we noticed something odd at PostNitro: a growing share of our users never open our editor anymore. They ask Claude, ChatGPT or an n8n flow to make the carousel for them, and the agent calls us in the background.

That got us building for agents properly. Your agent can now design an on-brand carousel, image post or short video and schedule it to LinkedIn, Instagram, TikTok, Threads, Facebook and Pinterest, all from one prompt.

Cheaper near-frontier models make it tempting to send more context on every call. For anything personal, that is the wrong place to spend the savings.

OpenAI announced GPT-6.1 Sol at DevDay on 2 October. Its API page lists $2.00 per million input tokens, $0.10 for cached input and $10.00 for output, with a 1,050,000-token context window. Prompts over 272K input tokens are billed at double the input rate and 1.5x output for the whole request. OpenAI says the model gives near-Astra intelligence at a fifth of Astra's price. That comparison is OpenAI's own claim about its own product, and I haven't seen an independent benchmark of it, so treat it as marketing until someone you trust has run it on your workload.

The same event added computer use to the Agents API, so an agent can operate software through its interface, available through the API and in Codex and ChatGPT Work on Pro 500 and Enterprise plans.

Here is what I think a small team should take from it. When the per-token price drops, the first instinct is to spend the savings on more context: send the whole history, the whole document, the whole account. The model can take a million tokens, so why not use them.

I m adding a short plan (a small challenge plus two or three steps) to the results page of an AI app. The endpoint that produces the result already chains several model calls: a safety check, the main generation, and two small side calls (a share caption and a memory note), all inside a 60-second function limit. Before adding a fifth call, I had Claude Code read the route and report how close it was to that limit.

The finding was uncomfortable. The two side calls had no timeout of their own. The SDK s default is one minute with two automatic retries, so one hung side call could run past the 60 seconds, and the platform would kill the whole request before the main result was saved. A code comment even claimed each call had its own timeout. It didn t.

What I changed: both side calls now get a 10-second timeout and no retries, and their failures are still swallowed as before. The new plan call goes through a helper that makes one attempt with an 8-second timeout and never throws. On a timeout, bad JSON, over-long text or a failed safety check it returns nothing, and the result is saved without a plan. It starts only after every safety gate, at the same moment as the side calls, so it adds little waiting time.

A test reads the route s source and checks that the plan call sits before the single database insert and that the side calls carry a timeout and zero retries, so a later edit can t quietly undo it.

Thanks to everyone who looked at AC Bridge on launch day. There is now a 66-second film showing what it does: https://youtu.be/amOOikdoy7Y

AC Bridge turns the smart lights you already own into status lights for Claude Code and Codex. The trial is 7 days with no card needed, then 2.99 a month.

Which agent state would you map to a light first?

hey PH Ola here, Specsight is launching today

we re all building faster than ever. a prompt, a few minutes, and a feature is live. it feels amazing!

hey PH Ola here, Specsight is launching today

AI made building software so much faster. but there s a problem nobody talks about: shipping faster is now costing companies money

Not just use AI for moodboards or a few shots but make it part of the normal production workflow.

What has to change first: quality, control, client trust, rights, pricing, or something else?