Jenny Park #BuildinPublic

Hey @X & Founders 👋 I'm looking to #connect with people interested in: 💻 SaaS 🚀 Startups 📣 Marketing 📈 Distribution ⚙️ Tech & automation 🤖 AI tools 📱 Mobile development 🌐 Web apps Let's grow together 🤝 Share what you're building! #letsconnect #buildinpublic

@Aditiiitm Retrieval frequency as a validation signal is smart. I like that it turns memory correctness into behavior, not just a static eval. Are you also tracking when a resolved memory causes a bad downstream action?

Great question Jenny — this is something we thought hard about. Three signals tell us if an auto-resolution was correct: 1. Retrieval validation If we resolved "user prefers Python" → "user prefers TypeScript" via UPDATE, and the TypeScript memory gets retrieved frequently in future sessions, that is implicit validation. Access patterns confirm the resolved memory is useful. 2. User correction rate We give users a consent dashboard where they can see every memory and correct anything wrong. If a user corrects a memory that was auto-resolved, that is an explicit signal the system got it wrong. We track this as a correction_after_auto_resolution event — it feeds back into confidence thresholds. 3. Clarification resolution accuracy For low-confidence conflicts we inject a clarification question into the next session ("Quick check — are you still using Python?"). When the user answers, we compare their answer to what the auto-resolver would have decided. This gives us a ground truth dataset over time. The honest answer: we are early on closing this feedback loop fully. The instrumentation is there but we have not had enough real users yet to tune the thresholds. That is the next frontier. What is your approach to validation on the memory correctness side?

Hey AI builders 👋 I'm building MemoryOS— shared memory + conflict resolution for AI agents so they can actually collaborate without losing context. If you're working on agents, multi-agent systems, or long-term memory — let's connect. What are you building right now? 👇 GitHub: [github.com/memengine/memo…]

@koustubh018 @X That’s a clever wedge. No-traffic-needed analysis makes it useful before launch too. Do you position it more as landing page QA, conversion audit, or product validation?

Purely AI analysis of the actual page experience - no real visitor data or session recordings involved. The AI agent navigates your page cold, clicks every CTA, and evaluates what felt unclear, confusing, or broken from a first-time visitor perspective. No traffic needed. Works before your first visitor ever lands.

Hey @X & Founders 👋 I'm looking to #connect with people interested in: 💻 SaaS 🚀 Startups 📣 Marketing 📈 Distribution ⚙️ Tech & automation 🤖 AI tools 📱 Mobile development 🌐 Web apps Let's grow together 🤝 Share what you're building! #letsconnect #buildinpublic

Validate your SaaS idea before you build. 3 free checks 👉jynlab.com/idea-analyzer

I’ll go first. I run ads and I’m building an AI ad generation pipeline because creating enough new ad ideas is always painful. I’m currently building it for myself, but I may open it to other founders running paid ads. What’s the hardest part of your ad creative workflow?

@Aladddinaliyev @X That sounds interesting. BYOI for multi-agent services is a strong angle. What’s the first workflow you’re trying to make reliable?

@createwithjenny @X hi i’m building BYOI system for multi agent services let’s connect

@TehnomasterS @X Nice to connect. AI + Android apps is a great space. Are you building consumer apps, internal tools, or something more agent-driven?

@createwithjenny @X Hey Jenny! 👋 As someone building AI & Android apps, I'm always looking to connect with experienced PMs and fellow builders. Let's definitely connect and grow together! 🤝🔥

@Aditiiitm That two-layer approach makes sense. I especially like escalating low-confidence or sensitive memory to the user. How are you measuring whether the automatic UPDATE/MERGE decision was actually correct over time?

Great question. We handle it in two layers: 1. System resolution — classify conflicts as UPDATE, MERGE, KEEP_BOTH, or REJECT. 2. User resolution — if confidence is low or the memory is sensitive, we ask the user to decide. Our view is that when the system isn't certain, the user should remain the source of truth. How are you handling agent disagreements today?

@koustubh018 @X This is very relevant for product validation. How are you identifying why visitors leave: behavior patterns, session recordings, or AI analysis of the page and traffic source?

@createwithjenny @X Hi Jenny, I'm building firsteyes.ai - Find out why first-time visitors leave your website before converting...

@birdpitchx @X This is cool!

@createwithjenny @X Building AI-powered social selling for X. Find relevant conversations, discover leads, and generate personalized reply drafts for organic outreach. No auto-posting. Human-reviewed replies only. birdpitch.com

@tanyaa_sen @X Hey I was a web developer in the past. Just a couple of thoughts: Since 3D is your main selling point, adding a quick demo clip of a page in motion would be the ultimate proof. If you can, make the background or header react to scrolling to really show off those motion capabil

@createwithjenny @X Yes landy.studio , Most landing pages look the same now: AI-generated, flat, forgettable. Landy generates $10K+ Worth 3D motion pages from one sentence, in ~10 min. The kind that actually turns traffic into customers 🔥

@mariyav4leva @X Let’s connect Mariya :-)

Alright cool my name is Mariya fractional CFO and founder of FounderFirst I've spent 13 years inside startups helping founders own the financial logic I've lasted: - 2 hypergrowth journeys, 4 pivots - 3 near-bankruptcies, 1 exit - 70+ founders in the portfolio since I write about building, founder journey, cash, growth, the messy middle, and the conversations founders avoid until they're too expensive to ignore My goal is to sprint to 10k by end of year if you're building and posting about it on X you're welcome to connect

@tusharthethe @X Hey I am also solving similar problems as yours! Nice to connect! Love the messaging here! It’s super clear on the 'why.' A quick visual or a 'How it works' step-by-step section would make the process feel even more tangible for new founders.

@createwithjenny @X Build startups with a problem-first approach.Reduce risk. Save time and money. startupmaya.com (WIP) Thoughts?

@ryan_zheng_work @X Looks clean! To make it clearer, consider: • The 'Why': Add one line on what problem it solves (e.g., 'Automate complex tasks locally'). • The 'How': A 5-sec demo GIF would instantly clarify 'Local-first Agent Teams'. • The 'What': Explicitly state the primary use case.

@createwithjenny @X Hey, I am building local native agent team application autobyteus.com youtu.be/xl3FdqWQaeI?is… youtu.be/QRMR6yOlYHw?is…

@koustubh018 @X This is cool Koustubh

@DNSAssistant_ @X This looks like a solid start! Try these: • Breaking the headline • Shortening the text • Adding a screenshot • Showing integrations

@createwithjenny @X Hey Jenny, Here is what I'm building dnsassistant.com

@DelavizDawodi @X This is cool! It might sound silly, but it would be better if you could explain more simply in the first sentence what this actually does on the landing page.

@createwithjenny @X Building Koven — a feedback, roadmap, and changelog platform that helps teams collect feature requests, prioritize ideas, and keep users updated as features progress. Would love your thoughts: koven.page

@JustJerry121 @UGCWithBritt I trust repeated behavior more than opinions: the same complaint from multiple users, people already paying for a workaround, or a problem that keeps showing up in reviews. What signal do you trust most?

@createwithjenny @UGCWithBritt Product validation is a solid place for Claude/Codex. Happy to connect here - curious how you decide which feedback to trust.

Founders and builders! I'm looking to connect with more: - growth hackers - marketers - AI builders - startup founders - vibe-coders - Claude/Codex power users - self-improvement maxis Or just cool people in general! Say hello 👋

@veeraj_ That’s a real problem. Discovery is painful even when the product is good. And yes, that’s exactly why I’m building jynlab: I got tired of guessing which ideas were worth building.

ShipCred is a directory I'm building to help founders showcase their projects and actually get discovered. It's a work in progress, stay tuned to find out :) btw love the jynlab idea, even I myself really struggled with finding a good idea (ranted about it some time ago on X as well😅)

Founders/builders, Building in public and want to meet more people doing the exact same. 🤝 Looking to connect with: ↳ Founders & indie hackers ↳ Full-stack & AI builders ↳ Micro-SaaS creators ↳ Anyone shipping something right now I'm building ShipCred - a dedicated directory for founders to showcase their projects, build an audience, and get discovered. What are you working on? Drop it below, let's connect and grow together. 🛠️👇