LAS G Concept

Insights · LAS AutoPost

What LAS AutoPost is and how it is built

By Lawal Ayo, founder of LAS G Concept · October 31, 2026 · 5 min read

Posting to social media by hand is repetitive: copy the headline, paste the link, adjust the length for each network, remember to do it again tomorrow. LAS AutoPost is the tool we built to take that repetitive work off a small team, without taking away the control that matters. This article explains what it is, the design decisions behind it, and where it stops. If you want the feature list and the plans, see the LAS AutoPost product page.

The idea in one paragraph

LAS AutoPost turns a blog, RSS feeds and long videos into scheduled posts for many channels. It writes captions in a brand's voice, places posts in a queue at sensible times and shows which links people actually click. It publishes to Telegram, Discord, Bluesky, Mastodon, Facebook Pages, Instagram, Threads, LinkedIn, X, YouTube, Pinterest, Tumblr and webhooks. It runs on Cloudflare Workers, the same family of infrastructure we use for our other sites.

Principle 1: automation must be predictable

A tool that posts in your name can do damage quickly if it misbehaves, so most of the engineering effort goes into restraint rather than features:

  • No duplicates. Before a post is queued, AutoPost computes a fingerprint from the text, link, image and destination. The same post cannot go to the same destination twice while one is waiting or sent. Feed items are recognised by both their identifier and their link, so a feed that changes its identifiers does not cause reposts.
  • Safe backlogs. Connecting an old blog does not post its archive. By default the backlog is recorded and only genuinely new items are posted.
  • Retries that behave. Temporary failures, such as rate limits and network errors, are retried with growing delays and a limit. Permanent failures are reported with the platform's message instead of being retried.
  • Brakes. Each destination has a daily cap. A workspace can be paused, there is a dry-run mode that publishes nothing, and an emergency stop exists for the platform owner.
  • Read-only tests. Testing a connection never posts anything.

Principle 2: AI suggests, it never blocks

AutoPost uses AI for captions, hashtags, grammar fixes, per-platform versions and highlights in videos. Every AI call is optional. If the AI is off, over its daily limit, too slow or returns something that fails the checks, AutoPost falls back to a plain template and records why. The checks matter: invented links are removed, numbers that do not appear in the source are rejected, banned words are refused and the emoji policy is enforced. Each workspace has a Brand Brain that stores voice, audience, rules, banned words, hashtags and a call to action, and these are applied both in the prompt and after the text is generated.

In the weekly summary of analytics, a validator rejects anything that invents numbers, promises growth or adds links, and a plain deterministic summary is used if the AI text fails. We would rather give a dull, true summary than an exciting wrong one.

Principle 3: honest analytics

Short links count clicks per post, destination and feed, and separate people from bots and link-preview crawlers, which would otherwise inflate the numbers. Only aggregates are stored: no IP addresses and no personal data. When two captions are tested against each other, a winner is declared only when there are enough human clicks, enough time and the difference passes a statistical test. Otherwise the answer is "no clear winner".

Principle 4: teams need roles

Once an agency, or a business with more than one person, uses a tool like this, the question changes from "can it post?" to "who is allowed to post, and who checked it?". AutoPost has five roles, an approval queue with an optional second approver, an audit log and one workspace per client. See approvals and client reports for how these work in practice.

Video clips, and why they need a helper program

One of the more unusual parts of AutoPost is the video studio. You upload a long video or paste a link, and AutoPost transcribes it, chooses between three and ten highlights from the timestamped transcript, and cuts short vertical clips with burned-in subtitles. Every clip edge is snapped to a real word time, so nothing is invented. Because the web platform AutoPost runs on cannot run ffmpeg, the cutting is done by a small separate program, the Video Worker, that runs on a computer you own and talks to the dashboard only through outbound secure requests. It also means the heavy work happens on your hardware.

What AutoPost does not do

It does not promise reach. It does not buy followers or engagement, and it does not bypass the rules of any network. Some networks, such as Facebook, Instagram, LinkedIn and YouTube, require an app review before an application can post for other people, and until that is approved a channel shows its status instead of pretending to work. It also does not replace a person with something to say: the best use of AutoPost is to remove repetitive work so that you can spend the time on content worth sharing.

Pricing, in naira and dollars

There is a Free plan with 3 destinations, 2 feeds and 100 posts a month. Pro is ₦12,000 or $9 a month, and Agency is ₦38,000 or $29 a month, with higher limits. Prices are per workspace. Because our users are often in Nigeria, naira billing is available alongside dollars.

Where to go next

Start with How to turn an RSS feed into scheduled social posts, then Plan a weekly posting schedule. When you are ready, open LAS AutoPost.


Written by Lawal Ayo. LAS G Concept is a Nigerian technology and media company. Questions? See the FAQ or contact us.

More from Insights