Your posts don't just appear.
Here's every moving part.
Tyashin runs a background engine that watches your schedule, pulls inspiration from four sources, decides what to write next, and publishes at exactly the right time. This page shows you every gear in that engine.
Settings
Four settings control everything
Configure these once in your LinkedIn settings. The engine does the rest.
When to post
You define exact times and a timezone for each posting slot. The engine pre-generates a post 2–3 hours before each slot fires.
How often
Choose daily, weekdays only, weekly, or a custom schedule. This decides which days your slots are active.
How many per day
Set a daily cap so the engine never over-posts on your behalf. Hard maximum is 4; recommended default is 1–2.
Auto-pilot mode
When ON: posts are generated and approved automatically. When OFF: every post waits for your manual review before scheduling.
Context assembly
Posts don't come from nowhere
Before the AI writes a single word, it assembles context from four sources.
Your brand profile & voice
Industry, target audience, tone of voice, vocabulary style, things to never say, controversy comfort level, admired profiles. This is the "who you are" context.
Your knowledge bank
Case studies, product docs, URLs, uploaded PDFs — any content you've added. The AI references this to ground posts in YOUR specific business details, not generic industry knowledge.
Live industry trends
Before writing, Tyashin searches the web for the latest news, market shifts, competitor moves, and events in your industry — so every post references what's actually happening right now, not outdated information.
Proven frameworks
Every post is built on a copywriting framework: Hook-Story-Offer, Problem-Agitate-Solve, and others. The AI chooses the right structure for the topic and funnel stage.
All four are combined into a single context window before Tyashin writes your post. The AI actively searches the web for current data, then blends it with your voice and business context. That's why it sounds like you — and references what's actually happening right now.
Post creation
A post can be born two ways
The engine can work silently in the background, or you can step in and steer.
Auto-generated
Engine runs on schedule
Up to 3 hours before your slot
Hourly cron checks each of your configured posting slots. If a slot is approaching with no post assigned and your daily limit isn't reached — it triggers generation.
AI assembles context
Your brand profile + knowledge bank + live web research (searches for current industry news, stats, and trends) + copywriting framework — all combined into a single context window.
Strategy is chosen
TOFU/MOFU/BOFU funnel stage + copywriting framework (e.g. HSO, PAS, Contrarian) are selected automatically based on your audience and recent post distribution.
Tyashin writes the post
Full output: hook, body content, reasoning chain, research brief with verifiable facts, first-comment suggestion, and hashtags.
Quality gates check
Hook length ≤ 140 chars, no external links in body, grounded in real data, different hook pattern from recent posts. Fails? Regenerates automatically (max 2 retries).
Post auto-approved and scheduled
Post is set to "Approved" and scheduled for the exact slot time. You receive an email with a preview link — review it, edit it, or delete it before it goes live.
Requires: Auto-pilot ON + posting slots configured
You in AI Studio
You steer the conversation
You open AI Studio
Pick from 17 theme cards (trend riding, hot takes, data insights, myth busters, etc.) or type your own prompt.
Chat with the AI
Back-and-forth conversation. Tyashin searches the web for real stats and news, then drafts in your voice. Refine, try different angles, iterate.
Post appears in chat
Full post with hook, body, "Why this works" explanation, research brief with verifiable sources, and first-comment suggestion.
You choose what to do
"Save to queue" — enters queue for the next available slot. "Schedule" — you pick exact date + time. "Rewrite" — new angle, same topic.
Post enters your queue
Queued for your next posting slot or the specific date/time you chose. Edit inline anytime before it publishes.
Works anytime — auto-pilot can be OFF
Approval gate
You decide how much control you keep
Every post passes through a gate. Where it goes depends on one toggle.
Status: Needs Review
Post waits in your queue
You read the content
View AI reasoning, see the hook breakdown
Edit inline if needed
Rewrite the hook, change the CTA, adjust tone
Click "Approve"
Status → Approved, enters publishing queue
Or click "Reject"
Gone forever, never publishes
Status: Approved immediately
No review step — post goes straight to queue
Assigned to next slot
The next available posting time is claimed
You receive a preview email
See what's coming — with a link to view it
Optional: manually unapprove
Catch it in time and pull it back if needed
Publishes at scheduled time
LinkedIn receives it at exactly the right moment
You can change this setting at any time from LinkedIn Settings. It takes effect on the next post generated.
Cron engine
What happens every hour
Tyashin's engine runs on the hour, every hour. Here's what it checks and does each time it fires.
Check usage thresholds
Every hourSends email alerts if you've hit 80% or 100% of your monthly AI allowance. Keeps you from hitting a wall unexpectedly.
Reset overage tiers
1st of month onlyOn the 1st at midnight UTC, clears any overage extensions from the previous month.
Auto-generate posts for upcoming slots
Every hourScans up to 3 hours ahead across all your posting slots. If a slot is approaching with no post assigned: generates one (with live web research), quality-checks it (retries up to 2×), auto-approves it, and schedules it for that slot. Skips if daily limit is reached or slot already filled.
Publish due posts
Every hourFinds all approved posts with a scheduled time ≤ now, plus any previously failed posts ready for retry. Publishes via LinkedIn API (batches of 10 in parallel). On success: marks published, posts your first comment if configured, sends confirmation email with a link to the live post. On failure: retries up to 3× with a 30-minute gap between attempts.
Check usage thresholds
Sends email alerts if you've hit 80% or 100% of your monthly AI allowance. Keeps you from hitting a wall unexpectedly.
Reset overage tiers
On the 1st at midnight UTC, clears any overage extensions from the previous month.
Auto-generate posts for upcoming slots
Scans up to 3 hours ahead across all your posting slots. If a slot is approaching with no post assigned: generates one (with live web research), quality-checks it (retries up to 2×), auto-approves it, and schedules it for that slot. Skips if daily limit is reached or slot already filled.
Publish due posts
Finds all approved posts with a scheduled time ≤ now, plus any previously failed posts ready for retry. Publishes via LinkedIn API (batches of 10 in parallel). On success: marks published, posts your first comment if configured, sends confirmation email with a link to the live post. On failure: retries up to 3× with a 30-minute gap between attempts.
Failure → Retry logic
If a post fails to publish, it doesn't disappear. The next hourly cron picks it up and retries — up to 3 attempts with a 30-minute gap between each. If all three fail, you get an email, the team gets a Slack alert, and the post is marked “Failed” so you can review and retry manually.
Now that you know how the engine works —
let's set it up for you.
Configure your posting slots, connect your LinkedIn account, and tell the AI about your business. The engine does the rest.