Pinned Tweet
Introducing Sumora
You tell it who you want to sell to. It finds prospects on LinkedIn and Instagram, researches them, spots buying signals, and ranks who is worth contacting. Then it writes the outreach, follows up, and can build a highly personalized lead magnet for every prospect.
why is ts so funny 😭😭
Here's a little timelapse of an investor call.
Honestly? I've always kinda loved them.
To see people learning about @trysumora and wrapping their head around how crazy the vision is.
The best part? The right type of investor insights ✨
Always a win.
(@ycombinator when?)
2
87
just submitted our 4th YC application.
rejected 3 times. probably getting rejected again. two 19 year olds from india with no connections don’t exactly fit the pattern.
but this time we have $3K MRR, 33% month-on-month growth, and we’re building shopify for AI agents.
applying anyway. would you let us in?
Your coding agent doesn't know about the bug in your exact version.
So I built landmine, and measured whether it actually helps.
I made 21 coding tasks from bugs filed after Claude's training cutoff. Each has a hidden test that the "obvious" code fails.
Claude Sonnet 5.5 on its own: 72/84 (86%)
With landmine's known-bugs list in CLAUDE.md: 83/84 (99%)
The part that surprised me: I also gave it web search, and separately the Context7 docs MCP. Across 84 sessions it never once used them to check for known bugs. It didn't think to look.
What landmine does:
→ reads your lockfile
→ finds open GitHub issues filed against the exact versions you pin
→ pulls the workaround from the issue thread
→ puts it where your agent reads first: AGENTS.md / CLAUDE.md, a pre-edit hook for 20 agents, or MCP
I also scanned 28 popular open-source repos. The median lockfile pins 395 open issues filed against its exact versions. I hand-checked 40 random ones: 30 were real.
Caveats: my own benchmark, only Claude Code tested, about 75% precision.
Free and open source. Try it: npx landmines scan
It's live on Product Hunt today. I'd love feedback, especially the false positives:
producthunt.com/products/lan…
18
19
232
No install needed:
• type any npm/PyPI package at landmine.tech
• or add the hosted MCP: claude mcp add --transport http landmine landmine.tech/mcp
Code, method and every benchmark transcript: github.com/devanshranjan10/l…
69
Contral crossed $1,000 on @AppSumo in under 6 hours
top performer badge, day one
two founders, no ad spend, one problem: AI writes code faster than we can read it
Contral teaches you every line as your agent writes it, then defense mode makes you explain it before it ships
lifetime deal, pay once. code contral10 = 10% off for new appsumo customers
contral.ai/appsumo/
1
3
82
"hey, i'd love to offer you a free audit"
everyone's received this. almost nobody replies
the audit isn't the problem. sending the same offer to everyone is
so at sumora our outbound agent does the work first:
- a breakdown of what their competitor posted last month
- 3 content ideas built for their account
- hooks written in their brand's voice
not "can i send you something?"
"here's something i made for you"
moved reply-to-booking up by at least 40% for our users
generic offers ask for attention. personalised ones earn it
1
90
your ICP tells you who could buy
signals tell you who might buy now
most outbound gets this backwards. build a list by title and company size, send the same message to 2,000 people, call 1% replies normal
but a list doesn't know timing
the founder who just raised. the agency hiring its first sales rep. the company whose competitor just launched
same ICP. completely different moment
that's what we built sumora's outbound agent around. find the signal first, then write for that moment
fewer messages. better timing. more meetings
5
1
96
when we started sumora we charged $49/mo
now every customer pays $500/mo
at $49 people treated it like a tool. tried it, poked around, drifted off
at $500 they treat it like a hire. they show up, tell us exactly who they want to reach, and care if it works
that pressure made the product better
cheap pricing doesn't build trust. it gets you taken less seriously
if it makes people real money, price it like it does
1
45
nobody wants AI agents
they want the meeting booked, the post published, the bug fixed, the ticket closed
the agent is just how it gets done
when we started selling @trysumora outbound agent, nobody asked about models or how many agents run under the hood. they asked one thing: will this get me more meetings
if you can't name the outcome, you don't have an agent
you have a chatbot with extra steps
4
74
Contral is live on AppSumo
AI writes code faster than we can read it. contral teaches you every line as your agent writes it, so the code you ship is actually yours
lifetime deal, pay once. starts at $79
code contral10 = 10% off for new AppSumo customers
appsumo.8odi.net/n4JVgx
1
78
AI writes code faster than we can read it
and most of us ship it anyway. it works until it breaks, and then nobody can explain what it actually does
that's why we built Contral
it's a teaching layer for any AI coding agent. while cursor, claude code, copilot, codex cli or our own agent writes your code, contral explains every function, pattern and decision as it happens
then defense mode asks you to explain the code in your own words before it ships. you get real feedback, not a pass or fail
a few other things it does:
- turns any repo into a step-by-step learning path
- build mode with plan mode, so you approve the plan before any file changes
- bring your own openai, anthropic or google keys
works in vs code, cursor, windsurf, antigravity, kilo code and vscodium, plus our own contral IDE for mac, windows and linux
tomorrow we're launching it on @AppSumo. lifetime deal, pay once, no subscription
goes live 13 oct, 8 AM CT (6:30 PM IST)
reply "contral" and i'll send you the link the moment it's live


















