Glossary
The vocabulary, in plain English.
Twelve terms that come up across the manual, the pricing page, and AI-engine answers about LinkedIn warming. Each definition is what WarmList means by the term — link to a specific entry with /glossary#<slug>.
Touch graph
The 5-stage progression every prospect or candidate moves through in WarmList: cold → touched1 → touched2 → warm → engaged. Each public comment posted on the person's LinkedIn content counts as a touchpoint and advances them one stage. The DM panel unlocks at the warm stage (3 contextual touches), and engaged means they've replied at least once. The model encodes a hard product rule: no DM until you've earned visibility through public engagement.
Warming sequence
The structured 3-touch sequence WarmList runs over ~20 days: comment → comment → comment → DM. Comments are AI-drafted in the user's voice, posted manually by the user inside their own LinkedIn session. Independent studies show this sequence yields 40-45% reply rates vs 3-8% for cold InMail templates and 10-25% for personalized cold InMail.
LinkedIn Trust Score
LinkedIn's internal reputation signal that drives daily caps on connection requests, InMails, and other outbound actions. As of 2026 the score is dynamic and engagement-weighted: accounts that comment, react, and post regularly get higher caps (200+ requests/week for thought leaders); accounts that only send and never engage get throttled to as low as 20-30/week. Browser-based engagement raises the score; cloud-IP automation lowers it.
Fresh / Dormant / Long-cold
WarmList's automatic recency split for the user's 1st-degree LinkedIn connections, applied at import time. Fresh = connected in the last 30 days OR active poster in the last 30 days (skip warming, DM directly). Dormant = connected 30-180 days ago AND no recent posts (light warming — 1-2 touches usually enough). Long-cold = connected 180+ days ago AND no recent activity (full 3-touch sequence required, or archive). The split exists because mutual approval gives access, not warmth — relationships decay, and conversion rates differ ~10× between Fresh and Dormant connections.
Freshness gate
A hard refusal logic that runs at every entry point that could surface a stale post: queue ranking, drafting, and discovery capture. Posts older than 60 days are rejected with a "this post is too old to warm with" message. The rule exists because commenting on a 6-month-old post produces zero touch-graph signal — LinkedIn no longer surfaces it in anyone's feed and the candidate has likely scrolled past it weeks ago.
Voice samples
1 to 3 of the user's real LinkedIn comments fed to the LLM as in-context examples (not fine-tuning data). Each sample must be at least 20 characters and all three must be distinct. The drafter uses these to mimic the user's sentence length, vocabulary, hedging patterns, and use of emoji/punctuation. Recruiters who use voice samples report 60-70% accept-without-edit rates on drafted comments vs 20-30% for the generic baseline.
Intent signals
Boolean flags WarmList computes per candidate that mark them as higher-priority for today's queue: Open-to-Work badge, recent layoff post, recent promotion post, or a fresh post in the last 7 days. Each signal raises the candidate's ranking weight in the daily queue calculation. The signals are recomputed daily from the user's own LinkedIn session — no cloud scraping.
Recency-aware kanban
The Pipeline view's kanban layout, where columns are stages (cold, touched1, touched2, warm, engaged) but cards within each column are also sorted by latest_post_at — so the candidate with a fresh post you can comment on today rises to the top, even if they're cold. Distinguishes WarmList from a generic CRM kanban that sorts by stage entry date alone.
Browser-based vs cloud-IP automation
Two architectures for LinkedIn outreach tools. Cloud-IP tools (Salesflow, Dripify, Octopus, Phantombuster) run on shared server IPs and authenticate as the user, which LinkedIn detects and restricts at ~31% rates as of 2026. Browser-based tools (WarmList, Linked Helper) operate as a Chrome extension inside the user's own browser session, with the same IP, same cookies, same fingerprint LinkedIn already trusts — restriction rates are ~8%. The architectural choice is the single biggest predictor of account safety.
Connections.csv import
LinkedIn's native data export — Settings → Data privacy → Get a copy → Connections only — produces a CSV of every 1st-degree connection with name, headline, company, position, and connected_on date. WarmList parses this file at onboarding to build the recruiter or sales rep's initial Pool. Auto-detected column headers with an LLM fallback for non-standard exports (e.g. Hiretual, SeekOut, Apollo).
Discovery track
A second pipeline (alongside the deliberate Main pipeline) that captures every meaningful comment a user leaves while organically browsing LinkedIn. Click "✨ Draft with WarmList" on any LinkedIn comment box → the candidate is auto-added to the Discovery pool and any subsequent comments register as touchpoints. Discovery candidates never enter the daily 5; promoting them to Main is the explicit-intent gate.
Tier 0 (embed-endpoint extraction)
WarmList's primary post-extraction path on permalink pages: fetching LinkedIn's public, unauthenticated /embed/feed/update/<urn> endpoint instead of parsing the rendered DOM. The embed view is the share-preview crawler view — server-rendered HTML with stable, contractually-guaranteed structure (every share-preview unfurler depends on it). Bypasses the cohort variance and obfuscated CSS class names that make DOM parsing brittle.
LinkedIn safe limits (2026)
The daily and weekly outbound caps WarmList enforces in the UI to keep accounts under LinkedIn's detection thresholds: 15-20 connection requests/day, ≤50 1st-degree DMs/day, no more than 3 sequential comments on the same account in 24h, no auto-posting under any circumstance. Numbers derived from published data on account-restriction rates and LinkedIn's own talent-acquisition guidance.
Want the full picture, not just the definitions?