Context and the gap
3D printing in 2021 was a maker's dream that demanded a lot of patience. Single-piece prints could run for hours; meaningful production runs took days. And in between every print on the bed, someone had to be physically present — to peel the part off, clean the surface, level the bed if needed, and queue the next job manually.
For hobbyists this was a quiet annoyance. For small businesses, makerspaces and educators trying to run printers as productive infrastructure, it was a serious throughput ceiling. The industry's existing automation answers were either expensive industrial systems or DIY hacks; nothing in between sat at a price and complexity that fit a desktop printer.
Single-job mindset
Most desktop printers shipped with no native concept of a print queue or remote control; "remote" usually meant a Raspberry Pi running Octoprint on the local network.
Operator-bound throughput
A printer needed a human in the loop after every print to clear the bed and start the next job — even when the machine itself could run all night.
Open ecosystem fragmentation
The desktop 3D printing market was dominated by open-firmware printers (mostly Marlin-based), but tooling around them was scattered, often hobby-grade, and rarely cloud-native.
The vision
We co-founded Spero 3D with a simple thesis: hands-off 3D printing should be a feature, not a luxury. The product had to do two things at once — physically clear the bed between prints, and give the operator a remote, cloud-based way to feed and monitor the queue.
That meant a tightly coupled hardware + software product. Either piece on its own would have been incremental; together, they reframed what a desktop printer could do unattended.
Hardware
An automated bed-clearing mechanism that could be retrofitted to existing Marlin-based printers without compromising print quality.
Software
A cloud platform where users could build print queues, monitor jobs from anywhere, and watch live camera feeds — collaboratively, across teams.
Distribution
Target both the maker community (open hardware, low-friction install) and small commercial print farms (queue management, multi-printer control).
Phase 1 — TÜBİTAK 1512 MVP
The first generation, developed under TÜBİTAK's 1512 program, validated the concept end-to-end. The hardware was a multi-piece chassis bolted under the printer's print bed, driven by its own DC motor and gearing. The software was a Python-based plugin for Octoprint that handled queue management, remote monitoring and the sequencing logic that triggered the bed-clearing mechanism after each successful print.
Octoprint plugin
Python + Jinja templates + JavaScript — print queues, remote monitoring, per-region print isolation during a job, nozzle/bed temperature control, live camera, timelapse and terminal access.
First-generation Reloader
Multi-piece chassis under the bed, dedicated DC motor and gearing, magnetic build sheet ejected and reset between prints.
Outcome
A working MVP that proved end-to-end automation, with enough live install hours to surface the next round of design problems.
Phase 2 — the 1507 rebuild
The 1512 MVP also showed the system's ceilings. The multi-piece mechanism's GT2 closed-loop belts were a constant supply-chain headache; the under-bed real estate was unforgiving for any belt-driven design; and Octoprint's hardware footprint (a Raspberry Pi 3B/4 per printer) had become disproportionately expensive after the global chip shortage. Under TÜBİTAK's 1507 program we rebuilt both halves of the product.
Hardware redesign
Instead of a self-driven mechanism, the second-generation Reloader piggy-backs on the printer's own motors and a custom end-of-print G-code script. A spring-steel build sheet, secured to a single-piece extension at the front of the bed, flexes upward when the Y-profile rises after a print finishes — flicking the part off and resetting itself in one motion. Far fewer moving parts, far more durability.
Software replatforming
We replaced the Raspberry Pi + Octoprint stack with a native print server running directly on an ESP32-S3 microcontroller. The S3's USB OTG capability let us speak to the printer's FTDI controller through a USB Virtual COM Port — a hardware path that, at the time, no one had implemented at this layer of the stack.
Plug-and-play Cloud Box
A device the size of a thick matchbox, with onboard camera, microSD storage and a WebSocket link to the cloud — replacing a Raspberry Pi setup that cost 5–10× more.
The cloud platform
The Cloud Box was the bridge; the cloud platform was where users actually lived. We built the backend on Go (chosen for its WebSocket and concurrency story) and the frontend on Vue.js + Quasar, shipping a unified experience across mobile and desktop. WebSocket was selected over HTTP polling, MQTT and RPC after benchmarking — its low-latency, persistent-connection model was the right fit for live print monitoring and camera streams.
Multi-printer dashboard
Queue, monitor and control multiple printers from a single account.
Print queues and libraries
Upload G-code once, organise into reusable lists, dispatch to any connected printer.
Live camera and timelapse
Streamed over WebSocket, optionally captured into per-print timelapse video.
Plug-and-play onboarding
Pair the Cloud Box from the mobile app, no networking knowledge required.
Collaborative use
Foundation for shared print farms — multiple users, queues that move across teams, resource sharing.
CES 2023 and the open-source bet
Spero 3D went to CES 2023 in Las Vegas as an exhibitor. The reception was warmer than we'd dared to hope — international beta sign-ups, retail conversations and the kind of validation a hardware startup measures in handshakes more than dashboards. Around the same time we made a deliberate strategic choice: open-source the hardware (now branded Autoflex), and keep the cloud platform proprietary. The thesis was simple — drive bottom-up adoption through the maker community, monetise through the cloud platform that the hardware naturally pulled people toward.
CES 2023 exhibitor
Live demo of the full Reloader → Cloud Box → mobile app loop on the floor in Las Vegas; international beta cohort recruited in person.
Open-source hardware (Autoflex)
Mechanical CAD, electronics and firmware released to the community for contribution and adaptation across printer models.
Closed cloud platform (Spero Cloud)
The SaaS layer where queue management, multi-printer control, live monitoring and team collaboration lived.
The market shift and the decision to wind down
While we were building the open-platform play, the market quietly tipped in another direction. Bambu Lab and the wave of closed-ecosystem desktop printers it represented arrived with vertically integrated automation already built into the printer itself — no add-on hardware, no third-party cloud, no community plugin. They could iterate at hardware-software integration speeds we couldn't match, because their stack was theirs end-to-end and ours had to retrofit a fragmented Marlin universe.
We made the call deliberately. With R&D capital tightening and the market timing shifted, sustaining the company would have meant either pivoting onto someone else's closed platform (giving up the thesis) or burning more capital chasing a head start that was no longer ours to take. Winding the project down was a hard call, but the right one.
Market signal
Bambu Lab and closed-ecosystem players landed with integrated automation, faster iteration cycles and a smoother out-of-box experience than any open-platform retrofit could match.
Capital reality
Continued R&D would have required follow-on funding in a tightening environment for hardware startups.
The call
Rather than dilute into a fight we couldn't win on the same terms, we wound product investment down — open-source artefacts and the cloud platform remain online for the community.
The product management lens
Looking back, Spero 3D was a product management exercise in three things stacked on top of each other.
0-to-1 across hardware + software
Leading discovery, scope, roadmap and delivery for a tightly coupled mechatronics + cloud + mobile product, where one half of the system was constantly redefining the other.
Funding-strategy match
Structuring the work to fit the rhythm of TÜBİTAK's 1512 (proof) and 1507 (productisation) programs, and managing reporting and audit alongside product execution.
Knowing when to stop
The hardest, most under-discussed product skill — reading market signal honestly, separating sunk-cost emotion from forward-looking decisions, and choosing not to keep funding a thesis the world had moved past.