One cold email got a reply from Evan Spiegel. Another led its recipient to publicly call it the worst he had received.
Both got attention. Only one screenshot shows a useful next step.
That is the problem with collecting “best cold emails”: memorable is not the same as effective.
A trigger is a fact about the company. A reason to contact it connects that fact to a problem you can credibly help with. These six examples make that connection explicit—and show when it is too weak to send anything.
The public examples below are attributed screenshots and posts; their limits are stated individually. The six copyable templates that follow are original, hypothetical teaching examples. None of our templates has a measured reply rate. They are message bodies, not complete sending instructions; use your real identity, appropriate recipient selection and opt-out handling. No outreach was sent to create this resource.
Download the editable worksheet · Free Markdown file; no signup.
Real cold emails: what the screenshots actually show
1. A short email earns a specific next step
David Ch shares an internship email from Niraj Pant to Evan Spiegel. The screenshot is dated April 14, 2014, not 2026. Niraj gives a compact account of his programming experience and asks how to pursue a summer internship. Spiegel’s visible reply appreciates the brevity and copies the lead recruiter.
Borrow: enough relevant background for the recipient to route the request. Do not infer: an offer, a hire or a repeatable reply rate. This is a recruiting example, not evidence that three bullets will sell your software.
Read the original post by @chhddavid
2. A product problem becomes a reason to reach out
Elliot Lindberg’s March 7, 2025 email asks about an internship at Lovable. He describes trying to build a client’s site with Lovable and Shopify, explains where the experience fell short, and offers to implement the integration during a defined internship period. His post says the email got him into Lovable at 17 and that the work had shipped.
Borrow: firsthand use → specific gap → concrete contribution. This is longer than Niraj’s email because the proposal needs context. The outcome is Elliot’s own report; the screenshot itself shows the outgoing message, not acceptance or shipping evidence. Neither example establishes an ideal word count.
Read the original post by @robiot
3. An operational deadline supplies a reason now
Suprava Sabat shares a Topo email that connects a stated Trigify shutdown date to a possible disruption in the recipient’s signals workflow. It offers a self-service test and credits before that deadline. The useful move is conditional: the pitch matters if the recipient depends on the affected workflow.
Borrow: verified change → affected workflow → relevant offer. The screenshot is evidence of what the sender wrote, not independent verification of the acquisition, shutdown date, product capabilities or recipient’s usage. Check those facts before adapting the structure. It shows no reply or conversion. We also would not turn its aside about links harming deliverability into a blanket rule.
Read the original post by @Suprva02
4. Attention bought at the expense of trust
Finn Mallery shares a recruiting pitch whose subject falsely announces a death in the recipient’s family. The body immediately walks it back, boasts about opens, then asks about a GTM role. Finn publicly condemns the message.
Reject: a fabricated personal emergency as a way to get an open. Even accepting the sender’s unverified open-rate claim would tell us nothing about qualified interest. The observable response here is public criticism. A subject line should help the right person recognize relevance, not trick them into fearing a tragedy.
Read the original post by @fin465
5. Diagnose the audience before polishing the prose
Tibo’s image summarizes advice attributed to Christina Gilbert’s YC appearance. Read the footer: the graphic’s headline framing is its creator’s, not hers. It is a secondary summary, not an original cold email or an official YC handout.
Its useful prompt for this worksheet is to check who you are contacting before iterating on copy. We are not adopting the graphic’s numerical reply benchmark, outreach quota or follow-up count as universal rules. Those need context that a shareable image cannot establish.
Read the original post by @tibo_maker
The common thread: make the recipient’s next decision easier. A relevant background, a concrete contribution or a verified change can do that. Shock alone cannot.
Six templates to adapt after checking the evidence
A company raised money. Your AI found the announcement. Your email says congratulations. Which part establishes that they need what you sell?
Use the following structures only after you can answer that question for your situation.
Choose the situation before the sentence
| Evidence you have | Example to use | Reason to skip |
|---|---|---|
| A current job listing names a relevant workflow | Hiring signal | The task is incidental or your sample does not address it |
| An unresolved public integration request | Integration demo | The request is closed or your connector cannot do the job |
| A company-announced migration | Implementation checklist | You only have an enrichment guess |
| A public problem and a working prototype | Design-partner request | You cannot state the time commitment or unfinished parts |
| A complementary resource for a partner’s readers | Resource partnership | It is a disguised link swap or an irrelevant brochure |
| An unanswered offer plus useful new information | Substantive follow-up | Refusal, opt-out, or nothing useful to add |
Before using any example, record five things: the observed fact and source date; the unverified inference; something useful you can deliver; the proposed message; and the condition that would make you stop. If the source is missing, leave the message unwritten.
1. A hiring signal with an operational question
Observed fact: a current company-owned job description assigns manual customer-report preparation to an operations hire. Save URL, exact relevant text and observation date.
Unverified inference: report preparation consumes significant time. A job listing alone does not establish hours lost, urgency or available budget.
Useful deliverable: an example report mapped to the advertised workflow, made with synthetic data.
Subject: the reporting workflow in your ops role
Your operations listing includes preparing customer reports. I build software that assembles those reports from [supported source]. I made a small example using synthetic data against the workflow in the listing. Would the sample be useful to whoever owns reporting?
Send only if you actually built the sample and support the named source. Skip if the listing is closed, reporting is incidental, or the sample cannot demonstrate relevant work. Never replace the question with an invented time-saving figure.
2. A public integration request
Observed fact: a company publishes a current request for an integration in its own public roadmap or forum. Record source/date and whether the request is still open.
Unverified inference: the company would adopt an outside solution. It may prefer to build internally.
Useful deliverable: a working connector demonstration with clear scope and constraints.
Subject: [system A] to [system B] request
I found your public request to connect [A] with [B]. We have a working connector for [specific operation]; it does not yet handle [important limitation]. I can send a short demonstration with sample data if the request is still active. Would that help?
Skip if you cannot perform the operation, the request is already resolved, or the evidence came from a private community you are not authorized to use. Do not call a prototype production-ready.
3. A real implementation decision
Observed fact: the company has publicly announced a migration that touches a workflow your product supports. Record the actual announcement, not an enrichment guess.
Unverified inference: the migration creates a gap your product should fill. A change of tool is not evidence of dissatisfaction.
Useful deliverable: a migration checklist for the specific integration, with known failure modes and test steps.
Subject: [workflow] during the [system] migration
Your announcement mentions moving [workflow] to [system]. We maintain an integration for that setup and have a checklist covering [concrete check] and [concrete check]. Would you like the checklist for your implementation team? It is usable without our product.
Skip if the checklist is a generic brochure, it depends on buying your product, or you cannot verify the migration. Do not promise an unbiased audit while concealing that you sell the solution.
4. Recruiting a design partner honestly
Observed fact: a company publicly describes a workflow problem relevant to a prototype you are building. A technical blog describing an old problem is not proof that it is still unresolved.
Unverified inference: the team has time to help develop a new product. Research participation and purchase intent are different.
Useful offer: a narrowly scoped design-partner session, with the time commitment, unfinished parts and what the participant receives stated upfront.
Subject: a prototype for [specific workflow]
Your post describes [specific workflow constraint]. I’m building an early prototype for that problem. It currently handles [working capability]; [missing capability] is unfinished. I’m looking for feedback in one 20-minute session using sample data. Would that be relevant, or is this already solved for your team?
Skip if you need production access, cannot demonstrate the stated capability, or intend to turn an advertised feedback session into an undisclosed sales demo. The 20-minute limit is a promise to honor, not a proven conversion tactic.
5. A distribution partnership with a real contribution
Observed fact: the prospective partner publishes resources for an audience your company can help with a distinct task. Audience overlap does not establish permission to contact their subscribers.
Useful offer: an original, usable asset that complements one named resource and can be assessed without a meeting.
Subject: a worksheet for your [resource]
Your guide covers [specific decision]. We made a worksheet that helps readers do [adjacent task], including [limitation]. Would you like to review it as a possible companion resource? There is no signup required to use it; we would agree attribution and placement before anything goes live.
Skip if the asset is gated contrary to that promise, merely repeats their guide, or the proposed partnership depends on sharing personal contact data without permission. Do not offer a backlink swap disguised as reader research.
6. A useful follow-up after an unanswered offer
Observed fact: you sent a specific offer and have not received a reply. Silence tells you nothing reliable about whether the person read it or why they did not respond.
Useful deliverable: one concrete clarification or new piece of evidence. If nothing useful changed, another message may not be justified.
Subject: Re: [original subject]
One clarification on the [asset] I mentioned: it works with [relevant constraint], and the example uses sample data. Here is [a permitted public preview]. If the problem is already handled, no action needed.
Skip after refusal or opt-out, when your previous claim was inaccurate, or when the only addition is “bumping this.” Do not manufacture urgency or claim the recipient opened the first email. This is an example of a substantive follow-up, not a recommendation for a fixed sequence length.
Before sending any example (all six situations)
Write down the evidence separately from the inference. Deliver the promised asset without adding an undisclosed meeting requirement. Use your actual role and product capability. Stop on a refusal or opt-out. A funding announcement by itself does not establish a buying need.
Measurement worksheet
Keep comparable audience and observation windows. Record delivered messages, distinct human replies, positive replies, meetings booked and opt-outs separately. Define each before a test. Deduplicate people; exclude automated replies from human replies. Use counts as well as rates. Tiny samples are diagnostic notes, not winning-template evidence. No results exist for these examples.
Verdict: the most useful output may be no email
Use these examples as structures, not scripts to paste into every account. Replace a placeholder only when the evidence supports it. If the promised demonstration, checklist or sample does not exist, build it before offering it.
For a founder with limited research time, a clear no-send decision avoids spending that time polishing a weak premise. That is an operating rationale, not a measured savings claim.
For the separate question of whether the channel earns its keep, use our cold-email evidence and reset guide. Keep channel economics there; use this page when you need to turn a specific, verified situation into a credible message. If the useful decision is not to send, a paid pilot program is another way to put a product in front of a buyer. A reply still has to land somewhere; Close vs Pipedrive is about calling and pipeline cost, not about the sentence.
Sources and method
The six templates and evidence-to-message rubric are original Cerrito synthesis. The five embedded posts were supplied by a reader and reviewed October 2, 2026, including each complete original-resolution image, not only the cropped embed preview. Public screenshots have not been independently authenticated. We distinguish visible replies, self-reported outcomes and commentary; none supplies a controlled performance comparison. The full-image links preserve access to lower sections that X may crop.
Our September 26 research reviewed existing RevenueFlow founder examples and Satellyte’s scenario templates. Scenario libraries already exist; the intended contribution here is pairing each message with evidence requirements and a decision not to send. We do not adopt either publisher’s performance claims for our examples.
Google’s sender guidance covers technical requirements for personal Gmail recipients. Those requirements are separate from whether the offer is relevant; passing authentication does not make the examples effective. No universal sending volume or sequence length is recommended here.
