Day 1: Broke student building an AI tool to help technical founders build in public
Hey IH,
Just started today. Posting here for accountability and feedback.
Who I am:
- Student, no prior startup experience
- $100 budget for the whole thing
- Self-taught technical (Next.js / TypeScript)
- Zero network, no co-founder, never sold anything to anyone before
What I'm building: SubKitt — an AI tool that reads your git commits and shipped features, and turns them into a week's worth of tweets in your voice. For technical solo founders who ship constantly but hate marketing.
Why this problem:
Kept seeing the same thing on X — technical founders saying "I ship so much but stay invisible because content is a second job." I'm one of them. Wanted to build the tool I'd use myself.
The plan:
- Week 1-2: talk to 20 technical founders, validate the pain
- Week 3-4: ship ugly MVP (Next.js + Supabase + Claude Haiku, ~$30/mo infra)
- Week 5+: try for first 3 paying customers at $19/mo
I'll post weekly updates here — revenue, mistakes, churn, everything. No polish.
Two questions for IH:
- If you're a technical founder building in public — what's the hardest part about keeping up with content?
- For anyone who's done this before: what's one thing you wish you'd known before starting your first SaaS with no money and no network?
Following along: @CheemaEdu on X
Trending on Indie Hackers
Respect for starting this openly — most people wait until things look good before sharing.
Your idea hits a real pain, but the risk isn’t building it — it’s whether founders actually use the content once it’s generated.
A lot of tools solve “what to post” but not “why it matters.”
Curious — how are you thinking about making this something founders stick with, not just try once and drop?
Good question — honestly the one I'm most worried about.
Watching existing content tools, "try once and drop" happens when the drafts feel generic. You get 5 decent tweets, realize they sound like a LinkedIn SEO bot, and churn by week 2.
My bet is that stickiness comes from two places:
Voice accuracy that actually compounds — the tool reads your past posts, learns which ones landed, adjusts. Leaving means losing that learned model.
Workflow depth — if it's reading your GitHub commits on a schedule and drafts are waiting for you Monday morning, it becomes part of your week, not a thing you open when you remember.
Both are hard to build well. Neither is guaranteed. Honestly might not know if it works until month 3-4 with real users.
Curious where you've seen content tools fail this — have you tried any that actually stuck?
Yeah — you’re thinking about the right failure point.
From what I’ve seen, most content tools fail not because output is bad, but because they don’t connect to a visible outcome.
Founders don’t really care about “good tweets” — they care about:
– replies
– inbound
– distribution actually compounding
If that loop isn’t clear, it always becomes “nice to have” → drop.
The GitHub → weekly draft flow is strong though — that’s closer to behavior than just generation.
One thing that might help:
tying output to some kind of feedback signal (even simple ones early on) so users feel progress, not just content.
Curious — are you planning to show any kind of performance/learning loop to the user, or keep it more lightweight for now?
This might be the most useful reframe I've gotten on the whole idea so far.
"Good tweets" vs "connected to outcomes" is a real distinction. You're right that without the feedback loop, it becomes a nice-to-have that dies when users get busy.
My honest answer on the performance loop: it's not in the v1 plan. I was going to ship "generate drafts" first, get 5-10 users, then layer in metrics once I had something working. But the risk of that approach is exactly what you're describing — if the first 60 days don't have a visible outcome, users churn before I ever get to ship v2.
I think the honest move might be: ship v1 with dead-simple tracking baked in — even just "you posted 4 drafts this week, the one about 2 got the most engagement." Not performance analytics yet, just pattern visibility. Let users start seeing which types of drafts land, even if SubKitt isn't optimizing on it yet.
Does that track with what you mean, or are you pointing at something more ambitious (like the tool adjusting future drafts based on what performed)?
Exactly — not analytics, just enough signal so it doesn’t feel blind.
If users see what worked → and next drafts improve → that’s the loop.
Otherwise it’s just another content generator.
Bigger risk though is positioning.
If it looks like “AI writing tool,” users won’t expect outcomes.
The ones that stick feel like:
→ growth / inbound systems, not content tools.
Curious — are you building a writing tool or a growth system?
This is the question I needed someone to force me to answer out loud.
Honest answer: what I've pitched so far is a writing tool. What I should be building is a growth system. The two have different prices, different users, and different product scope — and the gap between the two is exactly the gap between "nice AI toy" and "load-bearing part of a founder's week."
If I'm serious about the growth-system framing, v1 can't just be "drafts from commits." It has to at least show:
Which drafts drove engagement last week
What patterns are emerging in what lands for you specifically
Some signal of "you're growing because of this" even if it's directional
That's a different v1 than what I was planning. More work. Probably worth it.
Going to sit with this overnight before I decide. Thank you — this might be the most useful 30 minutes of thinking this project has had yet.
That’s the right direction.
Then the real leverage is making that obvious from day one — not something users “discover” after using it.
Because if it’s perceived as a writing tool upfront, they’ll use it like one → and churn like one.
The ones that win usually signal the outcome immediately, not the mechanism.
Worth pressure-testing how this reads cold to someone seeing it for the first time.
That's the test I wasn't going to run until it was too late — you just saved me 6 weeks of "why did this churn."
"Read cold to someone seeing it for the first time" is the honest benchmark. Right now my pitch is mechanism-first ("AI that turns commits into tweets"), which pre-frames it as a writing tool regardless of what I actually build underneath. The product and the positioning have to match or neither one works.
Still sleeping on the bigger call tomorrow, but I can feel the answer moving — not "is this a growth system?" but "how would a growth system read in one sentence to a founder who's never heard of it?" That's a harder question than I was ready to ask 3 hours ago.
Appreciate the push. Genuinely.
Then I’d make the call explicit:
If it doesn’t change distribution, it’s not in the product.
“commits → tweets” is mechanism.
“turn your weekly work into inbound” is outcome.
Those lead to completely different builds.
Most people try to bridge that gap later.
The ones that win decide it upfront and cut everything that doesn’t serve it.
If you read your current one-liner cold — it still sounds like a writing tool.
Fix that first before you build anything else.
Heard. The cold-read test on my current one-liner is brutal — it absolutely still reads as a writing tool.
Rewriting the positioning today before I write another line of code. Target is one sentence that leads with outcome ("weekly work → inbound") not mechanism ("commits → tweets"), and a v1 scope that only ships what serves the outcome.
Will post the new one-liner publicly when I lock it. Probably tonight. Feel free to destroy it if it still reads wrong.
Day 4 update — positioning v2.
After round 8 with @aryan_sinh, cutting "AI agent" and "weekly commits" from the headline — both were leaking mechanism.
New:
"You ship. SubKitt turns your work into inbound — so you get an audience without becoming a content person."
Keeping the "distribution layer for builders" framing (his sharper version) parked until I've earned the right to claim platform-layer positioning. Pre-MVP founders claiming infrastructure status reads aspirational.
One language-level sharpening at a time.
Day 3 update.
Spent the last 48 hours realizing my Day 1 pitch ("AI that turns commits into tweets") read as a writing tool to everyone who saw it. Thanks to @aryan_sinh and @StasyBashin, the positioning is sharper now:
You ship. SubKitt handles the distribution. An AI agent that turns your weekly commits into founder-led audience growth — so you can stop pretending you'll become a content creator.
Shifting from "writing tool" to "growth agent." Different product, different price, different user promise.
Next: build the v1 that actually matches this.
That’s awesome — really glad the discussion helped and that you were able to find clearer wording for the positioning. Sometimes a couple of the right phrases can change how the whole product is perceived 🙂
Curious to see how this evolves.
What stage are you at now — have you started the technical build, or are you still developing the idea?
This is getting into the part where most tools still quietly fail — even after getting the positioning right.
Even if you frame it as a growth system, there’s one gap I’ve seen:
Most tools show correlation, not causation.
“You posted X → got Y engagement”
But users still don’t feel: this happened because of the system
The ones that stick usually make the loop feel tighter, like:
→ “this specific post worked because of this pattern”
→ “next draft already reflects that”
So it feels less like tracking and more like progress compounding
Otherwise even growth tools drift back into “interesting dashboard” territory.
Curious — are you thinking about making that cause → effect loop explicit early, or letting users infer it?
This is the distinction I needed named. "Correlation tracking" vs "causation reinforcement" — most tools do the first and pretend it's the second. User feels it immediately.
Honest answer on v1: I can't ship a full causal-learning loop from day one (solo, building fast). But I can ship the explicit version of the loop in the interface — every weekly draft batch comes with one line that says "last week's best-performing post was X; this week's drafts lean toward that pattern." Not an ML model yet. Just human-visible reasoning.
That keeps the cause→effect loop explicit from day one, even if the actual learning is shallow behind the scenes. Users feel progress compounding even before the model can truly adapt.
Does that framing hold, or am I cheating the distinction?
That holds — and I think you’re already past the hard part.
At this point the bigger risk isn’t the loop, it’s how it reads in one line.
If it still even slightly sounds like a writing tool, you’ll attract the wrong users → and they’ll use it wrong.
Something like:
→ “turn your weekly work into inbound”
already feels closer to what you’re building than anything “content/AI” framed.
If you lock that direction, the name/brand should carry that outcome too — not the mechanism.
Curious what your current one-liner is reading like now.
Locked it earlier today, updated everywhere (X pinned, bio, GitHub, IH). Here's what's live:
"You ship. SubKitt handles the distribution. An AI agent that turns your weekly commits into founder-led audience growth — so you can stop pretending you'll become a content creator."
Honest read of it: "handles the distribution" + "founder-led audience growth" + "stop pretending" is doing the outcome-first work. "Weekly commits" is still mechanism-adjacent — but I left it because I think the concrete hook earns the right to live in the sentence (too abstract otherwise).
Open to being told it still reads wrong. If it does, want to know.
Close — but yeah, it’s still leaking “how”.
“weekly commits” + “AI agent” pulls it back into tool territory.
The strongest part is:
→ “you ship, it turns into inbound”
That’s the hook.
If someone feels they’ll get audience without becoming a content person, they don’t care how it works.
Right now it reads like:
growth tool for devs
It should feel more like:
default distribution layer for builders
Even the name — “SubKitt” still sounds like a tool, not that layer.
If this hits, people won’t describe it as software — they’ll describe it as:
“the thing that turns my work into users”
I’d push it further in that direction.
You're right — "AI agent" and "weekly commits" are load-bearing mechanism words. Going to test dropping them:
v2: "You ship. SubKitt turns your work into inbound — so you get an audience without becoming a content person."
Honest pushback on the "distribution layer for builders" framing though: I think that's Year-3 positioning, not Week-1. Stripe started as "7 lines of code for payments," not "payments infrastructure" — the platform framing came after they had the credibility to claim it. For a pre-MVP solo founder, "default distribution layer" reads like aspirational-speak from someone who hasn't shipped yet.
Current move: take the 80% of your feedback that sharpens the language now, park the 20% that reframes the category until I've earned the right to claim it.
Name change I'm also parking until I have 10+ users giving me feedback on it. One big rewrite at a time.
Does that split seem right, or am I under-claiming?
That split is right — you’re not under-claiming, you’re sequencing it.
If you try to sound like “distribution infrastructure” now, it reads like positioning theater. You haven’t earned that layer yet.
But one thing I’d push slightly:
Even at v1, it shouldn’t feel like a tool you use
It should feel like something that just happens once you ship
Your v2 is close — I’d just remove the last bit of friction:
“You ship. Your work turns into inbound.”
No name, no explanation — just the outcome.
If that line hits, then you earn the right to explain SubKitt.
Otherwise people evaluate before they feel it.
You're right — "You ship. Your work turns into inbound." is sharper as a pure hook. No brand, no mechanism, just outcome. That's landing-page-hero territory.
My current v2 is still naming SubKitt because it lives on my pinned tweet / bio / README, where people are already at my profile and need to know what they're looking at. Different job.
So the split I'm going with:
– Landing page hero (cold reader, first impression): "You ship. Your work turns into inbound."
– Tweet / bio / README (already at my profile): current v2 with SubKitt named.
If that reads right, I'll ship the landing page tomorrow at subkitt.com with the hero you just wrote. Thank you — round 9 of this has been worth more than most paid consulting.
That split is clean — exactly how I’d run it.
Hero = pure outcome
Profile = context + naming
Only thing I’d watch now:
Once that line is live, the next bottleneck won’t be clarity — it’ll be belief.
“You ship → inbound” is strong, but people will subconsciously ask:
why would this actually work for me?
So even a small proof layer early (example, pattern, or before/after) will make that line convert way harder.
Otherwise it risks feeling like a great promise without weight behind it.
Ship it — this is already ahead of most.
"Clarity first, belief second" is the pattern I'd have learned the slow way after the landing page went live and converted at 1%. Skipping that debugging cycle.
Concrete move for tomorrow's build: beneath the hero, one small proof-layer section showing the before/after of a real shipped thing. Probably using my own repo as the example — "this commit [screenshot] → this tweet [screenshot]." One example, not a gallery. Keeps the hero punchy and gives the skeptical reader something to weigh.
10 rounds with you in 4 days is genuinely more useful than most paid consulting. Not going to ask for round 11 — you've done more than earned the ship. I'll post the landing page tomorrow. If the conversion numbers are ugly in 2 weeks, I'll come back and report what I learned. Otherwise, I'll just show you the line when it's live.
Thank you.
Hey, this resonates a lot.
I’m a technical person (10+ years in dev), but this is my first time building a SaaS solo — doing it during maternity leave, with no funding and no team. Also trying to build in public around an MVP that came out of my own pain.
For me, the hardest part isn’t even writing content — it’s deciding what to talk about.
There are constant small changes every day, but no intuition yet for what’s actually valuable to share vs what’s just noise. That “editorial sense” doesn’t come naturally.
And yeah — the zero audience problem is real. I’ve never been a public-facing dev, no Twitter following, so I’m also experimenting with Indie Hackers as a starting point.
This distinction just reframed the whole thing for me. I was optimizing for "write content faster" — you just named a bigger problem: "know what's worth writing at all." Writing the words is maybe 20% of it. Deciding what deserves words is the other 80%.
The editorial sense thing feels like it's partly pattern-matching from watching what lands vs doesn't — which day-2 builders like us can't have yet by definition. We haven't posted enough to know. The uncomfortable fix is probably just posting more "noise" on purpose and seeing what actually resonates with people, then reverse-engineering the taste.
Maternity leave + solo SaaS + no audience is a level of hard most people underestimate. Following your arc.
I think for a lot of developers this whole “building in public” thing feels pretty unnatural at first. We’re used to building quietly, not narrating the process. Well, I don’t want to be a “blogger” 😀 and to improve that skill
But it does feel like the reality is shifting — if you want people to find what you’re building, you kind of have to show up and share, even if it’s messy at the beginning.
Good luck with your journey too — curious to see how it evolves.
This is the line I'll probably steal for SubKitt's positioning: "I don't want to be a blogger — I just want to improve that skill."
That's actually the user I'm building for. Not "aspiring content creator." Just "technical person who wants their work to be findable without turning into a marketer." Totally different product from the one most tools in this space assume.
Following your build too — dropping your handle / project name here if you want, so I can cheer along. Good luck with the launch arc; first solo SaaS on maternity leave is a level of respect most founder Twitter will never give.
https://www.indiehackers.com/StasyBashin - here you can see my project and posts. I just started here and will describe it all more soon! My MVP in progress, so public link is not available yet.
Just followed and read both your posts. The pediatrician-recommendation-in-a-5000-message-chat example is the kind of specific pain that makes me trust a founder's product way more than any pitch deck would.
"I don't want to optimize this process — I want to take it off my plate entirely" is actually the exact positioning I've been trying to find words for on SubKitt too. Different surface (GitHub commits vs Telegram chats), same shape: AI agent that quietly does the work instead of helping you do it faster.
Watching for the MVP. Happy to trade notes any time — solo agent builders is a narrower club than I thought 48 hours ago.
You’re really good at catching phrasing and pulling out lines from discussions that can turn into positioning or even the core philosophy of a product. That’s honestly a great skill.
To be honest, I haven’t been thinking about that much yet — I’m fully focused on the technical side right now, just trying to get the MVP into a working state. But it seems like something I should start paying more attention to.
And yeah, I also expected the community to be more active and give more feedback. But I guess that’s part of the process — when you’re just starting and still “a nobody,” it kind of works like this.
Thanks for the attention!
Thanks — honestly I'm just copying down sharp phrases from people like you and Aryan. Not generating them.
On the "nobody" thing — that feeling is real but I think it's more time-gated than judgment-gated. I saw something from Marc Lou where he said his first 6 months on X were basically posting into the void. Then momentum kicks in around month 3-4, not because the content got better, but because enough posts compound that one lands and brings people in.
Nothing to do about it except keep shipping and keep showing up here. Which you're doing. Rooting for the MVP ship.
Yeah, at the beginning it really feels like we have to support each other and collect loyal people and feedback piece by piece.
It’s definitely not a fast process — and it seems like everyone goes through this stage.
Wishing you patience — it’s something we all need here 🙂
And if you have a moment, feel free to check out my latest post https://www.indiehackers.com/post/telegram-is-broken-i-built-an-ai-to-fix-it-7c3ba7575c and vote in the poll —
would really appreciate your input.
Just voted and left a comment on the post. The "AI answers about the chat" framing (vs "filter the feed") is sharper than most Telegram tools — different product entirely. Rooting for the ship.
This comment was deleted 8 days ago.
This comment was deleted 11 days ago.