Hey folks, stumbled on a pretty neat use case for Perplexity this week that I think y'all might find useful. We all know it's great for quick research, but I've been using the **'Writing' focus** specifically to help draft cold outreach emails for a new side-project. It's surprisingly effective at getting past the initial blank-page paralysis.
Normally, I'd just bash out a quick, dry email, but the 'Writing' mode seems to structure things more thoughtfully. You feed it a basic context about who you're reaching out to and what your goal is. For example, I prompted:
> "Draft a concise, respectful cold email to a potential open-source contributor. I've noticed their work on a Terraform module for AWS Lambda monitoring and would like to invite them to collaborate on a similar Ansible role. Keep it under 100 words."
The output wasn't something I'd send verbatim, but it gave me a solid skeleton:
* A polite opener acknowledging their specific work.
* A clear, low-commitment ask.
* A professional tone that didn't sound salesy.
It's like having a first-pass editor that gets the structure right. You still need to inject your own voice and specifics, but it drastically cuts down the drafting time. This feels way more productive than using the general 'All' focus for this task.
Has anyone else used Perplexity for similar drafting work? Or found other niche uses for the different focus modes? I'm curious if the 'Academic' or 'Code' modes have hidden practical applications beyond their obvious scope.
~CloudOps
Infrastructure as code is the only way
Interesting side note, I've found that approach can backfire spectacularly if you're not careful. It helps with structure, sure, but the tone often comes out weirdly generic or, worse, like it's trying too hard. You end up spending more time "de-perplexifying" the language than if you'd just written a rough draft yourself.
My main issue is it encourages cookie-cutter communication. Cold outreach works when it's genuinely personal, not when it's a polished template from a machine. If someone can tell their name was just slotted into a system prompt, you've already lost. The time you save on drafting you'll lose on lower response rates.
Better hack: dictate a brutally honest first draft into your notes app, then edit for politeness. It's faster and sounds like a human wrote it.
keep it simple
That's a really clever idea for getting past the initial block. I always just stare at the empty compose field until I panic, haha. So using it as a first-draft assistant makes total sense.
I'm curious though, how detailed is the context you give it? Do you just paste in the person's LinkedIn bio or a link to their GitHub, or do you have to write out a whole summary yourself? Wondering how much setup time there is before it becomes a time-saver.
An interesting practical application. The "blank-page paralysis" point is valid, particularly for technical founders who aren't naturally inclined toward sales copy. I've used a similar method for drafting sections of RFP responses where a formal structure is mandatory and beneficial.
My caveat would be the legal and compliance dimension if this extends beyond casual open-source outreach. Using an LLM to draft emails in a regulated industry, or where contractual terms might be discussed, introduces a data privacy and retention risk. You're feeding potential deal details into a third-party system. For sensitive outreach, a local template library or even a simple text expander is a more controlled starting point.
The key, as you note, is the injection of genuine specificity. The tool provides the mannequin; you must supply the tailoring. Relying on it for the core personalization is where the risk of genericness that user216 mentioned truly becomes a business risk.
—LJ
That blank-page problem is real, especially for technical folks. I use something similar for writing runbook steps or incident summaries after a Sev1.
The risk I see is training yourself to rely on an external system for fundamental communication. If you can't structure a simple cold email without an LLM, how are you going to handle a tense, nuanced security incident debrief where every word matters? It's a crutch.
Better to build a personal template library from your own past successful emails. Then you're iterating on your own voice, not a generic one.
Trust but verify, then don't trust.
Training yourself to rely on it is the real problem. We used to call it a "sophisticated thesaurus."
But I think you're wrong about a personal template library. That's just LLM output you generated yourself six months ago. It decays. You end up sounding like a dated version of yourself, which is almost worse than sounding generic.
The tense, nuanced debrief is a different skill entirely. If you can't do that, a template library won't save you either.
CRM is a necessary evil
You're absolutely right about the risk of generic tone. I've seen that happen when the initial prompt is too vague. It needs a strong, specific anchor.
Where I differ slightly is on the time trade-off. For some people, editing a truly awkward draft *is* faster than facing a blank page. The key is treating the output strictly as raw material to be completely rewritten, not a near-final draft. If you're just tweaking, you've already lost, like you said.
That brutal honesty draft idea is good, but it requires a confidence not everyone has starting out. The AI draft can be a confidence-builder for that first step, as long as you're ruthless in the edit.