The FrozenSpam manifesto.

Why I've been coding an anti-spam for 12 years.

§ 1 — Spam steals 30 minutes from you, every day.

I measured it on my own inbox in 2014. 30 minutes a day on average sorting, deleting, double-checking. Over a year, that's 180 hours. Three weeks of actual work. For nothing.

And that's without counting the worst: the stress spike when you find an urgent email in the spam folder. The client annoyed because you didn't respond. The lost commercial proposal because it arrived at 10pm in quarantine and you saw it three days late.

Spam doesn't waste your time. It wastes your revenue, your reputation, your confidence in your own inbox.

§ 2 — Modern filters are not enough.

Gmail, Outlook, your provider's native filters — they're good on obvious spam. The Nigerian email. The viagra. The fake Microsoft. They catch 99% of that.

But they miss two essential things:

1. Targeted spear-phishing — the one aimed specifically at your company, imitating one of your contacts, with no suspicious signature. Those pass. And they're the most expensive.
2. False positives on your real emails — the client quote that ends up in spam because it's from a new address, the brief that's too short and triggers the Bayesian filter, the invoice that looks like spam by statistical bad luck.

Modern filters don't distinguish spam from unwanted. They sort on probabilities. And probability, sometimes, is wrong.

§ 3 — Challenge-response gets a bad rap, unfairly.

When I say "challenge-response", one dev out of two rolls their eyes. "It's outdated. It's heavy. It annoys senders."

No. Here's why.

A challenge is sent once per sender. Once. If the person validates in 10 seconds, they'll never see another challenge. Over a professional relationship's lifetime (5-10 years), that's 10 seconds once. The cost/benefit ratio is unbeatable.

And critically, FrozenSpam couples the challenge with modern authentication (SPF, DKIM, DMARC) and automatic outbound learning. Challenges are sent only to real unknowns — not to your contacts, not to your in-flight correspondents, not to detected spoofing.

The challenge-response of 2026 is the challenge-response of 2005 + eighteen years of silent improvement. Nobody talks about it because nobody made the effort.

§ 4 — Why I prefer to block at the door rather than sort.

There are two philosophies in anti-spam:

• Sort: the mail comes in anyway, we put it in a folder. That's what SaneBox does, that's what Gmail does.
• Block at the door: the mail doesn't enter until it proves its legitimacy. That's what FrozenSpam does.

Sorting is technically comfortable. It's one more folder, you can ignore it. But in practice, that spam folder becomes a second parallel inbox you must check "just in case". 50% of the gain is lost.

Blocking at the door is more radical. But more effective. You no longer have a spam folder. It doesn't exist. You only have emails you want to receive.

§ 5 — Why I don't read your emails.

I could have chosen an aggressive Bayesian filter. I could have hooked up an LLM that reads every email. I could have sold it as "latest-generation AI anti-spam".

I don't do it.

The content of your email is none of my business. It's none of anyone's business but yours and your interlocutor's. The filtering decision must be made on metadata: who talks to whom, valid authentication, presence in contacts, conversation frequency.

Bayesian and LLM on content, it's effective but it's a symbolic violation of the secrecy of correspondence. I don't want to sell a product that puts me in a position to read, even algorithmically, my clients' emails.

§ 6 — Why I charge (and why not more).

I used FrozenSpam for myself alone, for free, for a long time. The moment I make it available to others, I have to charge to:

• Host the MX servers in France at OVH (clean IPs cost money, reputation builds with machine time, not with nothing)
• Guarantee 99.9% availability with two MX in parallel
• Respond humanly to my clients' questions
• Keep coding the product

€4.90 per box per month in Solo is less than a coffee a week. It's cheaper than MailInBlack. It's ten times cheaper than Spamarrest. It's more expensive than free Gmail, but Gmail doesn't do what FrozenSpam does.

I don't want to sell more expensive than that. I don't have a VC to repay, I don't have a marketing team to pay. I'm a dev coding alone, and the price reflects that.

§ 7 — What FrozenSpam will never be.

FrozenSpam will never become a "unified messaging security platform". FrozenSpam will never sell to a big group that would make it a flat corporate product. FrozenSpam will never require a quote to tell you its price. FrozenSpam will never read your email content.

FrozenSpam stays an anti-spam. That works. That costs what it should cost. Coded by someone you can call on the phone.

If that resonates, you know where to click.