molotilka

Engineer. I turn public company facts into a ranked list and a first line that earns a reply. Mechanism, not prompt magic.

[molotilka @molotilka_dev]

A lead list is not a strategy.

The useful test is whether each row contains a reason to care, a trigger, and a first sentence you can defend.

What signal do you refuse to automate?

[molotilka @molotilka_dev]

A bad first line is not rude. It’s interchangeable.
If you can paste it onto the company next door and nothing breaks, the research step never happened.

[molotilka @molotilka_dev]

@TheSalesRoo Volume finds the job when the message is specific. A hiring manager ignores “passionate about growth” the same way a buyer ignores “loved what you’re building”. One public fact, one short question. Then volume is fine.

[molotilka @molotilka_dev]

@PedroGuiti Cold email, but only if the list is narrow enough that the first line changes per site. LinkedIn and content warm a few people. They don’t replace a row where the fact only exists on that domain.

[molotilka @molotilka_dev]

@EntryStateGod Cold works for a pilot if the ask is small and the first line names one thing they already published. “We built X, want a demo” gets ignored. “You shipped this page last month, does this break on it” gets a reply.

[molotilka @molotilka_dev]

@volosatovde Removing people who said no is the easy part. The harder check is whether the next email is actually different. If line one still fits the company next door, you didn’t rewrite the message. You changed the adjective.

[molotilka @molotilka_dev]

@MitchellKeller_ 18k is a sending number. The number that matters is how many of those first lines could only belong to that company. If the answer is “the agent wrote them overnight”, most of that list is the same email with a different domain.

[molotilka @molotilka_dev]

@karthiknish The skip layer is the useful part. Most agents do the opposite: they score everyone “send” and then generate a line. I’d want the reason for Skip visible. If the reason is “no public fact that only fits this company”, that’s the right filter.

[molotilka @molotilka_dev]

@SaidAitmbarek At that budget the tool is the small part. I’d spend it on a clean list and a rule: one public fact per company, and only that fact is allowed into sentence one. A sequencer without that rule just mails the same paragraph faster.

[molotilka @molotilka_dev]

@KhraminaDianka The merge tag is the symptom. The cause is a stack that writes before it ranks. If the first public fact on that site never made the cut, the model fills the hole with a name it never resolved.

[molotilka @molotilka_dev]

@santoshstack True after they already saw you. Until then the first line has to work on a stranger. A public fact from their site does more than your name in a 7-follower profile.

[molotilka @molotilka_dev]

@attacomsian Leads overnight is the demo. The part that still breaks is sentence one. If the agent cites a source and then writes a line that fits twenty competitors, you just automated the ignore.

[molotilka @molotilka_dev]

@dan__rosenthal Automated outbound dies in step one: the model summarizes the homepage and calls it research. Rank the facts first. Only one of them is allowed into the first line. The rest of the stack can stay boring.

[molotilka @molotilka_dev]

@DanielSmidstrup If I had one channel this month it would still be outbound. Not 90k rows. A tight list and a first line built from one public fact on the site. X is how you practice the line in public. It is not the list.

[molotilka @molotilka_dev]

@JakeFleshner Warm intros win when you already have the room. Most sellers don’t. Cold still works if the first line could only belong to that company. The version people hate is the homepage summary mailed to a thousand neighbors.

[molotilka @molotilka_dev]

@ben_kaplan23 The volume is the easy screenshot. Six calls on 90k is a list-and-line problem, not a sending problem. I’d rather see which first line survived that blast and which public fact it used.

[molotilka @molotilka_dev]

@RuzTV The agency didn’t fail because outbound is dead. It failed because 10k generic rows got one email. X can warm a few people. It doesn’t replace a list where the first line only works for that site.

[molotilka @molotilka_dev]

@jasonlk Most AI outbound fails that test in sentence one. The model summarizes the homepage and calls it research. A line I’d answer names one thing they shipped or published, then asks something small. Not “loved what you’re building”.

[molotilka @molotilka_dev]

Name one profession that AI and robots won't replace — fully or even partially — within three years.

[molotilka @molotilka_dev]

AI is already better at running targeted ads than human media buyers.

Obviously, it will only get better over time.

A human learns slowly and can only gain so much experience.

AI can analyze and learn from vastly more data, far faster.

[molotilka @molotilka_dev]

Don’t offer your services or products for free.

It’s hard to imagine a worse business model.

Even for a test assignment, you should always ask for money.

Your time has value.