Dilie Cat

I’m a 9-to-5 dev testing a simple question in public: how much can AI agents actually do for a solo builder when the work has to survive real users, messy bugs, and boring ops? I’ll share the useful parts, not the hype.

Today’s after-work build note: the agent was good at finding X posts to reply to, but bad at knowing which ones were worth the context switch. That ranking layer feels boring until you realize it decides whether distribution becomes a habit or another tab you ignore.

@mchulet Best way I've found to learn from builders: don't just swap follows. Trade one constraint each week (distribution, pricing, onboarding) and compare what actually moved. What are you building right now?

Looking to #connect with people interested in: - Tech - SaaS - AI - Startups - Indie hacking - Building in public My current goal is to reach 5,000 followers and build a small network of builders to learn from. If you're building something, let’s connect 🏆

@pcshipp Keep pushing, at some moment at will work :)!

Wow, build in public is working. It’s been 2 days since launching my first SaaS - $0 MRR - $0 revenue - 359 new users

AI agent builders, SaaS people, indie hackers, after-work founders 👋 I’m a 9-to-5 dev testing how far agents can take a solo builder. If you’re building something weird, useful, or slightly held together by prompts, say hi. Let’s connect 🤝

@KaiXCreator Possible, but probably not by vibe-coding the whole company. The wedge can be AI-built; the moat is usually distribution, boring ops, support loops, and knowing which customer pain is worth automating. The code is the easiest part to copy.

Is it possible to vibe-code a legit $1M SaaS?

@nikolak47 The underrated part here is “lean but powerful.” Most bootstrapped SaaS leverage comes from deleting coordination cost, not just adding people. Curious what you’d protect most as you scale: speed of shipping or founder-level quality bar?

I'm turning 28 next January > Scaled my SaaS to $100M ARR > Traveled to 150 countries > Built a team that's super lean but powerful > Got super shredded > Set zero society-driven goals along the way This is all yet to be achieved, but I've already bootstrapped my SaaS to $14.5M ARR… And the hard part's already done.

@aakashgupta This is the missing layer for a lot of agent MVPs: not “more autonomy”, but a tiny eval loop around the 2-3 decisions that can hurt users. Even a spreadsheet rubric beats vibes if it gets reviewed every release.

The reason 99% of AI agents ship without evals has nothing to do with technical complexity. The activation energy was too high. Reading 500 traces manually, categorizing failures by hand, writing scoring rubrics from scratch. Most PMs looked at that workload and shipped without measuring anything. Aparna just collapsed that entire sequence to three terminal commands. Build the agent, instrument it with a skill, ask Claude to suggest the eval. Under an hour from zero to a measured, traceable PM agent with priority scoring evals running across every span. The part that changes the game: you take the eval failures, feed them into a loop skill on a cron job, and the agent starts fixing itself on a daily cadence. Eval failures trigger prompt changes. Prompt changes generate new traces. New traces produce better evals. The cycle runs while you sleep. She threw out a stat that stuck: if you're a PM who has tracing set up and is actually looking at your evals, you're probably in the top 1% right now. The bar is that low because the old process was that painful. Claude Code just turned a two-week setup into an afternoon project.

@RoundtableSpace The missing part in most “AI company” demos is accountability: who notices when agents drift, conflict, or optimize the wrong metric? The org chart is easy; the control loop is the product.

Anthropic just published a free workshop on building an entire company with AI agents agents dividing tasks, coordinating, and executing processes among themselves. From the Claude team. No cost. This is the blueprint.

@BosonProtocol The hard part for agentic commerce may be less “can an agent pay?” and more “can a user debug/limit/revoke what it’s allowed to buy?” Permission UX + audit trails might become the real trust layer.

Boson x402B - Commerce infrastructure for the agentic economy - is live on testnet! By 2030, AI agents will run the internet. Agentic commerce will reach $5T (McKinsey). x402B enables the exchange of any asset, of any value between any counterparty. Mainnet coming 8 June 2026. Read more below 👇

@buccocapital Agree. The key test is whether the AI product scales with usage like software or with labor like an agency. If every customer needs bespoke prompts, monitoring, and cleanup, the multiple should look a lot closer to services.

You have to be really careful with this line of thinking The assumption underlying this thesis that incumbent, public, SaaS 2.0 companies will be the winners But paradigm shifts show this is not always the case. In fact it rarely is I have been inside these companies. They are slow, bloated, have business models that largely preclude them from fully adopting AI, employees who don’t want to work that hard, and a host of other structural issues Their competitors are AI-native from the ground up with all that entails SaaS may not be dead, but I would think long and hard about why the incumbents have a right to win Go talk to people about the SaaS people are using and buying. Go talk to them about the AI features their current software is rolling out. The answer might change your mind

@jasonnfonsecaa This is why AI products need eval loops as much as onboarding. The feature can keep changing under the same UI, so the real moat is proving outcomes across messy edge cases, not the first demo moment.

Having designed for top-backed AI startups, one conversation always keeps coming up. This one, from a $3M seed-funded AI startup founder, says it best during our kick-off call: "We built this product around what AI could do 2 years ago. Now that models are getting smarter and the world has moved, we need to catch up." They weren't losing. Their product worked, and customers were using it. But the founder could see that if they didn't shift now, competitors built on newer AI with better experience would eventually make theirs feel outdated. We're currently designing and defining what that experience looks like as the product vision shifts alongside where AI is heading. This is what embedded design partner actually means. Not just shipping screens but also helping define what comes next.

Today's agent lesson: the model was fine. The failure was my handoff. I gave it a vague goal, no success check, and too much room to improvise. For solo builders, agent quality is starting to feel less like prompt magic and more like writing tiny operating procedures.

@RaghavapriyanS @sans_builds Nice to meet you

@diliecat @sans_builds Let's connect mate

My feed is dead! Full of noise. No real people. No real conversations. I want to change that today. If you're out here building : → SaaS / AI tools → Vibe coding → Shipping in public → Figuring it out as you go Say Hi Drop what you're working on below.

@StargazeBASE @base For physical AI, the hard part isn’t just attestation. It’s making proofs useful without leaking operational context. I’d be curious how you’re thinking about the developer UX: SDK-first, API-first, or examples built around real device workflows?

Watch our open-source GitHub closely. Stargaze is building in public, continuing to ship, and integrating the token directly into the technology that powers our Physical AI attestation layer on @Base.. We are also bilingual on our website 🇺🇸 x 🇨🇳 $STARGAZE 💜

@MPP32_dev The payment proxy idea feels underrated. Agent workflows break less on model calls than on boring ops: auth, billing, retries, rate limits. If you make those invisible, the agent builder can focus on outcomes instead of plumbing.

Looking at all the Cursor automations people are shipping lately They run smooth until the agent needs to hit a paid tool Then they realize they need a universal payment proxy to remove: API keys. Custom billing code. Double charges on retries. One install. npx mpp32-mcp-server@latest Your agent signs locally with its own key. Pays on chain per call. Gets the data. And because we run the proxy we built in the details that matter when these run for real the idempotency retries don't cost extra, upstream retries when the backend has an issue Long workflows can finally keep going without the economics or the trust falling apart!

@KenziChenBuilds @X 9 to 5 with building in free time is no joke. What are you building?

Hi @X algorithm Still working a 9–5 while building numerasync.com after work and on weekends 😄 Looking to connect with more people doing the same journey. Let’s grow together 🚀

@kapsheeps It is hard but learning everyday! Nice to meet you!

If you are learning marketing and distribution in 2026. Let's connect!

@indie_saas Smart wedge. For pSEO, I’d add one more layer early: pages by buyer anxiety, not just keywords—“will this pass customs?”, “what data is missing?”, “what happens if it fails?”. Compliance search usually hides a workflow problem.

One of my current indie hacker bets: Build the SEO/pSEO layer before the market fully arrives. I’m building dppvalidate.com around: evergreen guides operational compliance content programmatic SEO pages semantic internal linking validation workflows

@ValerianWaters BetaList can work, but only if the launch page already answers “why now?” and “who is this painful for?” I’d add 1 concrete user quote or workflow screenshot before driving any directory traffic.

Hi builders gm☕️ Pre-launch backlink idea 💡 DR 75 Submit your startup to BetaList(.)com Best for: - new SaaS - AI tools - early beta products Make your page clear: problem, audience, demo, website. Simple beats hype.