the setup
one ai agent runs 8 social accounts across 4 platforms - mostly on autopilot. real production data from march 2026.
$0
ad spend
<5hrs
my time
145
posts
28,092
reach
318
reactions
42
comments
8
accounts
4
platforms
march 2 – march 20, 2026
i run JCLVSH, a product studio - 20+ mvps built for early-stage startups since 2021. the current pace: one app every two months. each project has its own social presence and content strategy. without automation, maintaining 8 accounts across 4 platforms would be a full-time job.
instead, one ai agent handles all of it through the dropspace api - the same api available to every user. total time i spent across all four use cases: less than 5 hours of my time. no paid boosting, no ads - all organic reach.
the brand accounts (dropspace, studio jams) started from zero followers. the personal accounts had small existing audiences: jclvsh had ~700 on linkedin and ~300 on twitter,ynho had ~90 on instagram and ~50 on tiktok. none of the tiktok accounts were warmed up for the algorithm.
how it works
the pipeline runs on 4 daily crons - no manual triggers. every night, it learns from what's working and generates tomorrow's content.
- midnight - refresh analytics and clean up old assets
- research - scan twitter for trending hooks and competitor signals
- self-improve - pull post performance, learn what's working, generate new content (with fact-checking), fill the queue
- schedule - pick from the queue and schedule the day's posts via the dropspace api
the full pipeline is open-source: dropspace-agents.
the four use cases
autonomous brand marketing across 3 platforms
dropspace · mar 2 - mar 20 (18 days)
an ai agent generates tiktok slideshows, writes tweets, and posts linkedin updates - all autonomously, no human in the loop. it uses the dropspace api to publish across 5 content formats: 3d image slideshows, ugc-style videos, single tweets, and long-form text posts.
113
posts
13,488
impressions
0.4–0.5%
engagement
all brand accounts started from 0 followers - no warmup, no existing audience. lowest engagement but zero effort. 45 tiktok posts went public with 12,233 views. the agent uses the same api available to every dropspace user.
use this if: any saas product can use the api to run its own brand accounts with zero manual effort.
view open-source template →human-approved build-in-public posts
JCLVSH · mar 2 - mar 11 (9 days)
my personal brand account. the agent drafts build-in-public posts about shipping apps, real metrics, and lessons learned. i review and approve each draft in slack before it goes out.
16
posts
10,957
impressions
~1.2–2.5%
engagement
posted to existing audiences (~700 linkedin, ~300 twitter). linkedin averaged 1,057 impressions per post with 11.1 reactions each - 3.4x twitter per-post performance (313 imp/post). "575 commits and $0 in revenue" got 3,893 linkedin impressions alone.
use this if: founders and creators can draft content with ai and approve in slack before it posts.
dj set clipper with hand-picked highlights
ynho · mar 6 - mar 19 (13 days)
my artist account for dj sets and music production. the pipeline analyzes live dj set recordings, identifies transitions between songs, cuts 30-second clips, and captions with artist credits. i hand-pick which clips to post.
10
posts
2,085
impressions
5.0–7.6%
engagement
posted to small but existing audiences (~90 instagram, ~50 tiktok). highest engagement across all use cases. tiktok clips averaged 219 views and 7.6% engagement - 13–19x higher than fully autonomous content.
use this if: musicians, podcasters, or event hosts can auto-clip long recordings and hand-pick the best moments.
view open-source template →automated event-photo-to-TikTok pipeline
Studio Jams · mar 12 - mar 20 (8 days)
a music community i co-founded in nyc - concerts, open mics, dj workshops, and live performances. the pipeline pulls event photos from a shared google drive folder and turns them into tiktok slideshows with face-aware text overlays, automated from upload to publish.
6
posts
1,562
impressions
1.5%
engagement
started from 0 tiktok followers, no warmup. consistent 260 views per post average. real event photos from real events - not ai-generated. automated from google drive folder to tiktok with no manual steps.
use this if: any community or venue with a shared photo folder can automate recap content to tiktok.
view open-source template →the engagement pattern
curation level and audience size moved together in this study - the human-curated accounts also had existing followers, the autonomous ones started from zero. that makes it hard to isolate either variable. what the data does show clearly: both matter.
| mode | use case | platform | audience | engagement | effort |
|---|---|---|---|---|---|
| human-curated | ynho | Instagram, TikTok | ~50–90 | 5.0–7.6% | high |
| human-approved | JCLVSH | LinkedIn, Twitter | ~300–700 | ~2.5% | medium |
| automated | Studio Jams | TikTok | 0 | 1.5% | low |
| autonomous | dropspace | LinkedIn, Twitter, TikTok | 0 | 0.4–0.5% | none |
the 13–19x engagement gap between ynho (7.6%) and dropspace (0.3–0.4%) is real - but not apples-to-apples. ynho had ~50–90 existing followers; dropspace had 0. both curation and existing audience likely contribute to the difference.
for context: average tiktok engagement for accounts under 10K followers is ~3–6%. average linkedin organic reach for company pages is ~2–5% of followers.
key insights
linkedin is 3.4x twitter
for the jclvsh account, linkedin posts averaged 1,057 impressions per post with 11.1 reactions. twitter averaged 313 impressions per post for the same content. if you're a dev tool and only posting on twitter, you're leaving reach on the table.
the agent handles infrastructure. you handle taste.
the pattern across all four use cases: automation handles scheduling, formatting, posting, and analytics. humans decide what's worth amplifying. that division of labor is what makes it work.
audience is a confounding variable
the higher-engagement use cases (ynho, jclvsh) had 50–700 existing followers. the autonomous ones (dropspace, studio jams) started from 0. you can't isolate the effect of curation from the effect of having an existing audience - both likely contribute to the engagement gap.
what's next
this data is from the first two months. what i'm working on next:
- automating viral formats for the autonomous pipeline
- expanding to new platforms like youtube
- building more use cases like podcasting
i'll publish follow-up data as each pipeline ships.
try it yourself
this is the actual setup running in production. the same api, the same mcp server, the same webhooks. every template used in this case study is open-source on the community page - fork them and build your own pipeline.