AI Catalyst C3·Core Session - Week 13·2:50:00

Session 25: Product Hunt Launch Playbook — Engineering as Marketing, Rules That Get You Banned, and a Real Launch Scheduled Live

Harshit Instructor - runs the whole session solo: platform tour, his own profile setup, an /apps launcher page built by his Codex-in-Hermes agent, an agent-generated launch brief, and the actual Product Hunt submission of 'Grain Studio' scheduled for Tue Sept 1 · Shivani Co-host - opens (explains why Fireflies note-taking stays on per learner feedback), closes with CSAT

▶local recording · LMS lesson

The short version

Product Hunt puts 'a usable product in front of product-minded early adopters' - feedback, visits, social proof, founding customers. It 'cannot create Product Market Fit', fix onboarding or make a weak offer compelling. Philosophy: 'just launch' - MVP, beta or waitlist, disclosed as such (Cursor shipped 0.01; Microduck took $500 pre-orders).
Paul's question - doesn't launching before completion do the product a disservice? - gets the strategic answer: it depends on what you want true seven days after launch; a disclosed beta collects testers and founding customers, and the same product can be launched again ('10 different versions'). Session 26 then shows the post-launch consequences.
The rules: DO ask people to visit, try, share, comment; disclose affiliation. DON'T ask for upvotes, run vote giveaways, buy traffic or bots - and 'if your growth consultant is using prohibited tactics, you will be held responsible.' Ranking is pure upvotes, reset daily: ~300-500 on a quiet day, 500-700 competitive, 700-1,000 to be sure of #1.
The strategy has a name: ENGINEERING AS MARKETING - 1752VC gives away a pitch-deck analyzer that filters founders and markets the firm. Harshit copies it: an /apps page on his own domain hosting micro-tools, each launched separately, all funnelling back - 'clients approach me... YC companies... ask me to build out their entire playbook.'
The live launch: Google sign-in, full profile (photo, headline, bio, links, interests, work-email, GitHub/LinkedIn - never Apple, which reset his profile), keep the handle consistent everywhere and NOT hidden from search engines; three skills + repo + URL -> a PDF brief (positioning, listing copy, maker comment, 3-day runway, 24-hour go/no-go, contingency); the form (tagline <=60 chars, tags, screenshots incl. dashboard not just output, 240x240 icon, shout-outs, free/no promo, skip investors); scheduled Tue Sept 1 12:01 AM PST; embed badge in the site header. Tuesday or Wednesday, three days to a week out - never same day.

Outskill's post-read

The written recap Outskill published after the session (Session_25_Product_Hunt_Launch_Playbook_Post_Read.docx). Where it is sharper than the live session, those points are here; the full concepts below are from the recording.

'A Product Hunt launch does not create product-market fit by itself' - it reaches early adopters, drives visits, collects proof and starts founding-customer relationships; it cannot fix onboarding or a generic offer.
Maker profile first: same photo and handle as LinkedIn, headline, links, displayed product, interests, verification - consistency across platforms is also an SEO consideration.
Reverse-engineer: scrape past launches (Codex + Apify + Firecrawl) -> examine how they were presented -> identify patterns -> structure the findings -> a repeatable launch plan.
Brief the AI with the product, GitHub repo, the Product Hunt Launch / Launch Plan skills, launch context and research; get positioning, timing, assets, outreach, shareable content and a PDF plan.
Position by the specific difference, not the category (Grain Studio: 'tactile image transformation happening privately in the browser, with a visible result in seconds').
Schedule days ahead as an upcoming launch, prepare for about a week, Tuesday or Wednesday, and ask 'what do we want to be true seven days after the launch?' Ask for visits and feedback, disclose affiliation, never buy votes.

Its headings Product Hunt as a go-to-market channel · Launching is a repeatable skill · Choose what is ready · Build the maker profile first · The maker is part of the launch · Reverse-engineer before you launch · Give the AI the right context · Positioning · Build the listing · Timing · The pre-launch window · Launch-day mechanics · Connect your website · Launch ethics · Micro-launches · A product can support the main business · Other launch platforms · The complete playbook · The role of Codex

At a glance, three clicks deep

Skim here first: the closed row is the glance, open is the study card with the key points and timestamps, and the ↓ link drops to that concept's full write-up below.

01What Product Hunt can and cannot do for youProduct Hunt = free daily discovery/traction platform for early adopters;0:10:00›

Product Hunt = free daily discovery/traction platform for early adopters; not PMF, not onboarding, not an accelerator; alternatives HN, There's An AI For That, AppSumo.

Gives traction, feedback, social proof, founding customers (0:14)

Cannot create PMF or fix onboarding or a weak offer (0:14)

Free to launch; ad-funded; AngelList/Wellfound 2016 (0:32)

YC = accelerator (~6% via SAFE); Hacker News is its launch community (0:32-0:34)

There's An AI For That, AppSumo as analogues (0:35)

↓ Full write-up of this concept

02Paul's question: is launching before completion a disservice?Launch MVPs/betas openly labelled;0:29:00›

Launch MVPs/betas openly labelled; define the desired state seven days after launch; relaunch per version; the risk is post-launch neglect, not early exposure.

Paul: doesn't launching pre-completion do a disservice? -> depends on strategy; disclose beta and launch (0:29-0:31)

Precedents: Cursor 0.01, Microduck pre-orders (0:27-0:31)

Beta testers are part of the point (Deborah) (0:31)

'What do we want to be true seven days after launch' (0:15)

Same product can be launched repeatedly as versions ship (0:18)

↓ Full write-up of this concept

03The rules that get you banned, and what #1 actually costs in upvotesAsk for engagement not votes;0:16:00›

Ask for engagement not votes; disclose; no giveaways/paid traffic/bots; daily upvote ranking (300-1,000 for #1); 24-hour homepage window.

Do: visits/tries/shares/comments, disclose, real relationships (0:17)

Don't: ask for upvotes, giveaways, paid traffic, bots (0:17)

You are responsible for your growth consultant's tactics (2:51)

Ranking = upvotes, daily reset; 24 h on homepage (0:25, 2:50)

#1 needs ~300-500 / 500-700 / 700-1,000 by competitiveness (2:31-2:32)

↓ Full write-up of this concept

04Engineering as marketing: free micro-tools as the funnelBuild free micro-tools, launch each on Product Hunt, host them under one domain (/apps), funnel traffic to…1:17:00›

Build free micro-tools, launch each on Product Hunt, host them under one domain (/apps), funnel traffic to the core offer; sellable as a service.

1752VC analyzer = filter + marketing (1:27-1:29)

'Engineering as marketing' named (1:29)

/apps page on his own domain for micro-tools (1:53-2:00)

A real client service for YC companies (1:20)

Many micro-products, one domain, one launch each (2:54)

↓ Full write-up of this concept

05Profile setup: Google sign-in, uniform handle, never Apple, never hiddenGoogle SSO;0:42:00›

Google SSO; full profile; consistent handle; visible to search; own name as maker; avoid Apple linking.

Google sign-in first, links after (0:58)

Apple linking reset the profile live - avoid (1:15-1:16)

Uniform handle across platforms for SEO; don't hide from search (1:02, 1:16)

Colons rejected in links; underscore/hyphen for taken handles (1:03, 1:09)

Your name as maker, not the agency; any team email works (0:57, 1:05)

↓ Full write-up of this concept

06From three skills to a scheduled launch: the agent brief and the submission formSkills + repo + URL -> agent PDF brief -> override date (Tue/Wed, 3-7 days out) -> submission form -> sched…2:16:00›

Skills + repo + URL -> agent PDF brief -> override date (Tue/Wed, 3-7 days out) -> submission form -> schedule -> embed badge.

Agent brief: positioning, copy, maker comment, 3-day runway, go/no-go, contingency (2:24-2:26)

Agent's month-out date rejected; Tue/Wed, 3-7 days out (2:21-2:23, 2:58)

Tagline <=60 chars; tags; screenshots incl. dashboard; 240x240 icon (2:34-2:40)

Shout-outs, free/no promo, skip investors (2:41-2:43)

Scheduled Tue Sept 1 12:01 AM PST; newsletter Wed Sept 2 (2:50, 3:00)

Embed badge in header; embeds are website-only (2:52-2:57)

↓ Full write-up of this concept

The concepts in full

What Product Hunt can and cannot do for you

0:10:00

It is a room full of the people most likely to try your thing on day one. It is not a machine that makes them come back.

Can: traction on launch day, feedback from product-minded early adopters, visits, social proof, relationships - the founding customers a new product needs. Cannot: create product-market fit, repair weak onboarding, make a generic offer compelling, guarantee a returning audience. The platform is free to launch on (it monetizes with ads; acquired by AngelList/Wellfound in 2016) and is not an accelerator - Y Combinator takes ~6% via a SAFE; its launch community is Hacker News ('Show HN'). Related venues: There's An AI For That (AI-specific; even Salesforce used it), AppSumo.

Why it matters

Sets expectations before Paul's own question - the platform's job is discovery, and the disservice risk lives in onboarding, not in launching.

Paul's question: is launching before completion a disservice?

0:29:00

Paul asked the question everyone with a half-built product is thinking. The answer reframes 'finished' as 'what do we want true seven days after launch?'

Harshit's answer, relayed from Paul's chat question: it depends on strategy. If you have an MVP or beta, launch it - and say so. Physical products take pre-orders for things that do not exist yet (Microduck, ~$500, 'all over Twitter yesterday'); Cursor launched a 0.01 before its real v1. A disclosed beta on Product Hunt is precisely how you get beta testers (Deborah's follow-up) and founding customers; the listing can be launched again as versions ship. The organising question, stated earlier: 'what do we want to be true seven days after launch' - if the answer is 'fifty people trying it and telling us what breaks', an unfinished product is the right product to launch.

The honest counterweight comes the next day in Session 26: a live launch summons bot cold-emails, paid-upvote sellers and a PLG audit that found no return loop - the disservice risk is real, but it sits in what you do after launch, not in launching.

Why it matters

Directly Paul's concern for his own tools; the seven-days-after framing is the decision test he can reuse.

For your projects

Paul asked this. For Paul's Lovable/Supabase tools the practical version is: label it beta, set the seven-day goal, and put the subscribe-to-download funnel from Session 26 in place before the launch date.

This idea elsewhereextendsSession 26's PLG audit is what 'after launch' looks like for the product launched here appliesSession 4's micro-prototypes are exactly the kind of thing to launch unfinished

The rules that get you banned, and what #1 actually costs in upvotes

0:16:00

Ask for a visit, a try, a comment - never for an upvote. The platform's own dashboard says you are liable for whatever your growth consultant does.

Do: ask people to visit, try, share, give feedback, comment; disclose your affiliation; use real relationships. Don't: directly ask for upvotes, run vote giveaways, use paid traffic or bots - growth-hack spam Harshit discourages for first-timers even though 'many companies' do it. Product Hunt's policy text: 'If your growth consultant is using prohibited tactics, you will be held responsible.' Mechanics: listing shows ratings, tags, images, optional video, reviews, upvotes, follows, collections, analytics; ranking for Product of the Day is upvote count, reset daily as a new cohort launches; products stay on the homepage 24 hours then archive. Thresholds at recording: 300-500 quiet day, 500-700 competitive, 700-1,000 very competitive.

Why it matters

Session 26 shows the $1/upvote sellers arriving in the inbox within hours - this is the rule they are asking you to break.

This idea elsewhereextendsThe paid-upvote offers Session 26 refuses are the 'don't' list arriving by email

Engineering as marketing: free micro-tools as the funnel

1:17:00

A VC firm gave away a pitch-deck analyzer. Every founder who used it qualified themselves - and learned the firm's name.

1752VC's Pitch Deck Analyzer (launching that day) filters applicants while marketing the firm. 'The term for this is called... engineering as marketing': build small free tools, launch each on Product Hunt, route the audience back to the main product, site or agency. Harshit applies it live - an /apps launcher page on harshit.com (built by his Codex agent inside the Hermes VPS, header/footer nav, deployed via Azure Static Web App after Vercel/GitHub Actions trouble) hosting Grain Studio and future micro-tools - and confirms it is a paid service: YC clients ask him to 'build out their entire playbook and do the launch.' Ahmad's question - ten apps on one site, one launch each? - 'yes, that's the strategy.' Reverse-engineer competitors' launches with Apify/Firecrawl; skills.sh already has Product Hunt strategy skills.

Why it matters

For an agency this is the productization path Catalyst keeps circling: tools as lead magnets, Product Hunt as distribution.

This idea elsewhereextendsSession 26 rebuilds this /apps page into a marketplace and settles subdomain vs path appliesSession 24's audience-capture is where the funnel lands

Profile setup: Google sign-in, uniform handle, never Apple, never hidden

0:42:00

He connected his Apple account and his whole profile vanished. 'Please don't connect with Apple.'

Sign in with Google, then add links. Complete everything: photo, headline, bio, GitHub, Twitter, website, Instagram, LinkedIn, newsletter, interest tags, work-email verification, GitHub and LinkedIn connections. Handle unavailable? underscore or hyphen. Colons are rejected in link fields. Keep your handle and name identical across Product Hunt, LinkedIn and everywhere - 'all these SEO companies... are now focusing on uniformity' - and do NOT enable 'hide my profile from search engines'. Use your own name as maker, not the agency (Suraj); any associated person's email works (Palak). The platform is product-centric - makers are visible inside the listing (Mohan).

Why it matters

The uniformity rule is the same knowledge-panel logic Session 26 states for six months of consistency.

This idea elsewhereparallelSession 26's knowledge-panel uniformity rule

From three skills to a scheduled launch: the agent brief and the submission form

how-to2:16:00

The agent's first plan launched a month out. 'Way too long' for an indie tool - he moved it to Tuesday.

Feed the agent three skills.sh Product Hunt skills plus the GitHub repo and the live URL; it returns a PDF brief - positioning statement, listing copy, maker comment, distribution templates, measurement plan, a 3-day runway, a 24-hour go/no-go and a one-week contingency (Sept 8). Override the timing: Tuesday or Wednesday, three days to a week out, never same day - anticipation, social sharing, and the 'launching tomorrow' listing. Form: name; tagline <=60 chars ('make tactile image textures entirely in your browser'); tags Design Tool / Web App / Open Source; description; several screenshots showing the dashboard and before/after, not just output; 240x240 PNG icon (agent-made); maker = solo; shout-outs to Codex, Vercel, GitHub with agent-drafted lines; pricing free, no promo ('it's bootstrapped'); skip 'connect with investors'; schedule Tue Sept 1 2026 12:01 AM PST. Then prompt the agent to add the Product Hunt upvote embed to the site header (embeds are website-only; share links for social). Newsletter launch planned Wed Sept 2.

Do it in this order

01Give the agent the three Product Hunt skills, the repo and the live URL; ask for the launch brief PDF
02Set the date yourself: next Tuesday or Wednesday, 3-7 days away
03Product Hunt -> Submit: name, tagline (<=60), tags, description from the brief
04Upload screenshots (dashboard + before/after) and a 240x240 icon
05Makers, shout-outs, pricing (free / no promo), skip investors
06Schedule at 12:01 AM PST on launch day
07Add the upvote embed to your site header; prepare share links for social

Why it matters

A complete, copyable launch checklist - and the reason Session 26 opens with a live launch to analyse.

This idea elsewhereparallelSession 12's GTM launchboard is the planning artefact this brief automates

Tools referenced

Tool	Coverage	Moment	Context
Product Hunt	demonstrated	0:20	Tour, profile, submission, scheduling, embed
Codex	demonstrated	1:17	'Terra Lite' mode; built /apps page, brief, icon, embed
skills.sh	demonstrated	1:19	Three Product Hunt strategy skills
GitHub	demonstrated	2:05	Repo push; Actions trouble; shout-out
Azure Static Web Apps	demonstrated	2:12	Final deployment target ('Vercel, but on Microsoft')
Hermes	explained	1:23	His agent VPS ('Jarvis brain') hosting Codex
Screenpipe	explained	1:24	Mac-only screen recorder for a personal company brain; free via GitHub
Vercel	mentioned	2:17	Original host; abandoned mid-demo
Apify	mentioned	1:17	Scrape competitor launches
Firecrawl	mentioned	1:17	Alongside Apify for launch research
Fireflies	mentioned	0:07	Kept on for learners who revise from notes
Remotion	mentioned	0:19	Possible launch-video tool
Hyperframes	mentioned	0:19	Possible launch-video tool

Session materials

Archived locally on V: — click to open. Companion pages link to the LMS.

vtt

Week 13 Session 1 - Transcript.vtt

203 KB

Action items

Resources mentioned

Resources

docLaunch brief PDF, the three Product Hunt skills, and the playbook
docQ&A roll

Extraction notes

This page was built from an auto-generated transcript, which garbles product and people's names. Those were corrected silently in everything above and logged here for transparency. The warnings flag claims that were true on the recording day but change fast.

Transcript corrections applied

The transcript says	The trainer actually means
Green Studio	Grain Studio (his image-texture tool)
1252VC	1752VC
Terra / Terra Lite	a Codex model/mode label - uncertain
army's brain	Hermes brain / Jarvis brain (his agent VPS)
Appify	Apify
Whisper Flow	Wispr Flow
xkaliDraw	Excalidraw
Microduct	Microduck
NA10market.com	an n8n-themed domain he flipped for ~$2,000 - spelling unclear

True on recording day — verify before relying

Upvote thresholds (300-1,000) and the 24-hour homepage window are August-2026 observations
Grain Studio launch Sept 1 2026 and newsletter Sept 2 - see Session 26 for results
Product Hunt acquisition attribution (AngelList/Wellfound, 2016) stated loosely - verify
OpenClaw 'acquired by OpenAI' and 'highest paid product' are as told
Codex 'Terra' mode naming is as heard and may be a garbled model label