Skip to content
Get your invite
Runs inside Slack · Private beta

Stop scrolling X.
Start a /discover.

The agent searches X, scores every post 1–10, and drops a ranked digest in your Slack DM. You triage in minutes — it learns from every tap.

// sign in with Google · free during beta · no credit card

~2 hours/day · doomscrolling for a handful of good threads

5 min/day · triage a ranked digest

The problem

X is two hours of noise hiding five good threads.

Your feed blends sports, drama, and outrage with the few expert conversations where your take actually counts. Finding them is the work. Noiseless does that work — and only that work: no sports, no entertainment, just the industry threads that match your skills, scored 1–10.

How the loop works

Three beats, then it runs itself.

01

Digest

~5 posts, scored 1–10, land in your Slack DM each morning — curated overnight while you slept.

02

Decide

Tap 👍 Like, 👎 Dislike, or 🙈 Skip. One tap each. The whole review takes about five minutes.

03↻ repeats daily

Learn

Every tap retrains your ranker in the background. Tomorrow's digest is sharper than today's.

Discovery engine

It reads for meaning, not keywords.

The agent understands your profile — your stack, your domain, your goals — then scores every candidate thread 1–10 on how much your voice would matter there. You see the score before you spend a second.

Ranked results · today 5 matches
@systems_lead Event sourcing without the regret — what we'd do differently 9/10
@backend_ali Outbox pattern at 10k req/s: Postgres or Kafka? 8/10
@dx_maya most AI agents are a for-loop + a system prompt — show me state 6/10
@devrel_max Is the CLI dead for internal tooling? 5/10
@noise_acct my weekend vlog is up 🎥 — below threshold, not sent 3/10

// posts scoring below 5/10 are filtered out before your digest is sent

Semantic ranking

Ranked by meaning and relevance, not hashtags.

1–10 score up front

Triage before you read a single word.

Never repeats

Deduped across every run, automatically.

The learning loop

The longer you use it, the less it wastes your time.

No sliders. No preference forms. Your taps are the training data — the ranker retrains automatically in the background.

Week 1

42%

like-rate

Week 4

78%

like-rate

like_rate: 42% → 78% ✓

// directional figures from the builder's own beta usage — not a customer average. Your mileage will vary.

Step one · one time · ~5 min

Tell it who you are. Once.

Build a quick profile — your skills, a couple links to your work, and optionally connect X or GitHub so it can learn your voice. That's the whole setup. Discovery starts that night.

Build your profile

PythonDistributed SystemsLLM infra
github.com/you · yourblog.dev
Connect X / GitHub to enrich
Profile ready — first digest within 24h
1

Add your skills & domains

What you do and what you want to be known for.

2

Link your work

GitHub, a blog, your X — optional, but it sharpens matching.

Profile ready

Your first digest arrives within 24 hours.

"Can't I just prompt this?"

You could. Here's the part that's actually hard.

A prompt + the X API
Noiseless
starts fresh every session
Memory across runs
remembers every tap, compounds your taste
static prompt, same results
Ranking that improves
ranker retrains on your feedback
you wire the DB + Slack + cron
Dedup & delivery
never repeats, lands in Slack at 8am
you watch the API bill
Cost control
per-user caps + full audit trail

A prompt is a great tool for one session. An agent is a system that gets sharper while you sleep.

Why this exists

“I was losing two hours a day scrolling X to find five conversations worth my time. So I built the thing that does the scrolling for me.”

Noiseless is a real private-beta product, built by one engineer who got tired of the feed. There's no growth team and no fake reviews here — just a tool I use every morning and am now opening to a small group. The mechanism is exactly what's on this page: it discovers expert threads, you tap, it learns. If that sounds useful, the beta's open.

— the builder · Noiseless

Built by one engineer Testing with a small group Free during private beta

Controls & trust

Transparent cost. Your data stays yours.

Daily budget

$0.12 / $1.00

Monthly budget

$3.40 / $30.00

Data isolation

Row-level security, multi-tenant isolation, every API call audited. We never log your tokens or private content.

FAQ

Questions a skeptic would ask.

Finds expert X threads matching your profile, scores each one 1–10, and sends ~5 a day to your Slack DM. You tap 👍 Like / 👎 Dislike / 🙈 Skip; it learns.

The opposite. It targets professional and technical conversations — backend, infra, devtools, your domain — and explicitly excludes sports and entertainment.

Every Like, Dislike, or Skip retrains your personal ranker in the background — no sliders, no setup. Like-rate measurably improves over the first weeks of use.

It's the relevance score the agent assigns each post for you. Anything below 5 is filtered out; your digest shows the strongest matches first.

Yes. Row-level-security multi-tenant isolation, encrypted tokens, and a full audit trail. Your data is never shared between users, and tokens are never logged.

Free during the private beta. Per-user daily and monthly budget caps mean it can't run up a surprise bill.

Private beta · invite-only

Get five hours of your week back.

Five minutes reviewing a sharp digest beats two hours scrolling a noisy feed. Build your profile once — the agent handles the rest, and gets better every day.

Start with Google

// free during beta · no credit card · your data stays yours