The current curriculum looks polished because we ran it badly twice. This page is the honest story of how the program evolved — what the first cohort tried that didn't work, what a later cohort refined, and how the current model documented across the rest of the library got to its present shape. Some of what changed was big (program length, field site, NDVI introduction). Some was small but consequential (the 22-item pre-flight checklist came from one mid-mission radio-loss event in an earlier cohort; the Saturday-session option came from one trainee in the first cohort who would have made it with one more weekend). Most of what works now is here because something earlier didn't.
The rest of the Lumipad library — curriculum, business, the recruit page, the six detailed week pages — describes the program as it currently runs. It reads as a coherent system because, after two completed cohorts, it mostly is one. This page is the work that produced that system: the choices we made early that turned out wrong, the choices we changed mid-stream, and the patterns that survived two cohorts of trainees and became the cohort default model.
Designed for three audiences. Partner organisations evaluating Lumipad use this to understand what kind of organisation we are — one that learns and updates, not one that pretends the first version was correct. Internal staff training new instructors use this to understand why things are the way they are; the reasoning behind a decision is more durable than the decision itself. Future cohort 03+ trainees curious about the program's origins use this as context — the graduate community they're joining started smaller and rougher, and the cohort program they're completing is the third version, not the first.
A note on honesty. Some of what follows is uncomfortable to write. The first cohort had a near-miss prop strike. A later cohort had a paid-survey conflict that almost ended a partner-cooperative relationship. One trainee in the first cohort was held back at end of Week 4 and didn't graduate — a decision we still think about. We're writing this honestly because pretending the first cohort was a smooth pilot would make the next cohort's instructors over-confident. The current curriculum is robust because we earned every line of it the hard way.
Quick visual reference. Each row is a programmatic choice; the columns are how that choice has evolved across the three cohorts. The cohort 03 column is the cohort default current model documented in the curriculum.
Each row above is expanded below in narrative form. Numbers are accurate to the cohort program's internal records; trainee identities are anonymised throughout.
The first cohort ran from March to May 2023 with 4 trainees recruited through cooperative network referrals — three from cacao cooperatives in the Davao region, one from an abaca-growing cooperative further north. The cohort was deliberately small because nothing about the program had been tested yet; the team wanted enough scale to learn from but not so much that mistakes would harm too many trainees. The 8-week program ran in a workshop in central Davao with a vacant lot 200m away serving as the flying area. There was no simulator practice, no NDVI rig, no Mission Planner, and no formal paid survey component. The drone was a Matek F411 build with unmodified Pi Camera and generic tri-blade props.
An 8-week curriculum. The original program was structured as 8 weeks: 2 weeks of theory and soldering, 2 weeks of build, 1 week of configuration, 2 weeks of flight practice, 1 week of business. The thinking was that a longer program would produce better-prepared graduates. In practice, weeks 1–2 had too much theory and not enough building; trainees lost momentum. Weeks 5–6 of flight practice were too long without a goal — flying for the sake of flying without a survey to deliver became repetitive. Trainee fatigue was visible by week 7.
RGB photography rather than NDVI. The first cohort used unmodified Pi Cameras producing standard RGB aerial photography. The thinking was that NDVI was too complex to introduce in the first cohort and standard photos would still demonstrate value to cooperatives. This turned out to be wrong. Cooperatives looked at the RGB images and said, in effect, "I can see this from a tall ladder." The differentiated value of drone-based survey work — seeing crop health beyond what visible light reveals — wasn't being delivered. Two of the three cooperatives that engaged early graduates for paid work after graduation didn't come back for a second survey.
A vacant lot for flight practice. The Davao team had access to a vacant lot 200m from the workshop. The lot was free, accessible on foot from the workshop, and had no overhead obstacles. It also had several problems we didn't anticipate: it was bordered by residential housing on three sides; neighbours complained about drone noise during week 4 hover practice; the lot was unusable for two weeks during a particularly wet spell because the ground became soft and trainees couldn't set up the LiPo bench safely; and most importantly, there was no farmer relationship — trainees never flew over actual cacao or coffee until they did paid work post-graduation, which meant their first encounters with real crops happened with no instructor support.
An untethered first hover. The first cohort had no tether protocol. Each trainee's first hover was simply: take off, hover at ~1m, descend, land. The instructor stood at the LiPo bench with their TX in standby. One trainee's first hover went sideways — a yaw drift the trainee didn't correct, drone moved laterally toward the perimeter, instructor took TX control and landed it. Drone was ~2m from a non-trainee bystander when it landed. No injury, no equipment damage, but the team realised that night that "no injury so far" was not the same as "safe protocol." The tether protocol was developed over the next weekend and used for the remaining trainees' first hovers; it became standard for later cohorts.
Pass / fail assessment. The first cohort used a pass / fail philosophy. At end of Week 4 (which then included the build and basic flight assessment, before the curriculum was restructured), one trainee was held back: they had not achieved a clean tethered hover and the team judged they weren't ready for the field work that followed. The trainee was offered enrollment in a later cohort but declined; they returned to their home cooperative without graduating. The team still thinks about this decision. The current Saturday-session and optional Week 7 supervised practice options exist directly because of this trainee — we now believe they would have made it with one more weekend of one-on-one practice. The pass / fail philosophy was retired after the first cohort.
No formal paid survey. Early graduates were referred to interested cooperatives but the program didn't broker the first paid survey. The thinking was that this was beyond the cohort program's scope; graduates would handle their own first jobs. This produced a hard transition: graduates left the program with technical skills but no first invoice, no signed contract, no client relationship handed over. Of the three early graduates, one became a working pilot in the 18 months following graduation; the other two stalled at the "quoting first job" stage. The later-cohort paid-survey component (and the cohort-shared escrow at graduation) were direct responses to this.
Small cohort size let us see every trainee. Four trainees meant the lead instructor knew every trainee's build status, every solder joint, every flight tendency by name. This produced strong individual outcomes for the three who graduated; it also surfaced problems early — the held-back trainee was identified as struggling by week 3, not week 4. The current cohort default 12-trainee size requires more instructor capacity but tries to preserve this individual-attention discipline through smaller cohort cells (cohort default 6 trainees per cell with a dedicated instructor each).
The cohort 'cell' structure that survives today. The first cohort was 4 trainees and 1 lead instructor. When the team scaled to a later cohort (8 trainees), they organised into two cells of 4 trainees each, each with its own instructor. The structure stuck. The current cohort has two cells of 6. The instructor-trainee ratio (1:6 maximum) is a first-cohort inheritance.
The cooperative-referral recruitment pipeline. Three of the four first-cohort trainees came through cacao cooperative leadership referrals — cooperative managers nominating young members with the right hands-on aptitude. This recruitment channel produced the highest-yield cohort recruits the program has ever had; the current recruit.html still leans on cooperative referrals and program graduates-as-recruiters as the primary channels.
Workshop infrastructure decisions that paid off. The original first-cohort workshop (built out before the cohort started) made several infrastructure choices that proved durable: dedicated soldering stations with proper ventilation, a fireproof bench for battery work, separate clean/dirty zones, and a workshop wall for cohort photos and reference materials. None of these have changed across three cohorts.
Specific changes a later cohort made because of the first cohort
The first cohort graduated in May 2023. A later cohort started in December 2023 — a deliberate gap. The team needed time to redesign the curriculum, build the NDVI rig protocol, establish the partner farm relationship, and (less visibly) recover from the held-back-trainee decision. The team also used the gap to follow up with early graduates and learn what their first 6 months post-graduation actually looked like; many of the changes documented above came directly from those follow-up conversations rather than from the cohort experience itself.
One observation worth noting. Most of the durable improvements between earlier cohorts came from following up with graduates after they'd left the program, not from in-cohort observation. The instructor team's view of what was working was incomplete; the graduates' lived experience filled in the rest. The current 30/60/90-day graduate community checkpoints exist partly to make this kind of feedback loop continuous rather than periodic.
A later cohort ran from December 2023 to February 2024 with 8 trainees organised into two cells of 4. The program was cut to 6 weeks, the field site moved to the partner cacao farm, NDVI rigs were introduced on every build, the tether protocol became standard, and Saturday make-up sessions were available. The drone was upgraded to the SpeedyBee F405 V4 (the current cohort default) with the modified Pi Camera NDVI rig. Mission Planner (INAV) was introduced as the planning tool. Most of the current curriculum dates from this cohort; the structural choices made in the first cohort are what subsequent cohorts have refined rather than replaced.
The 6-week curriculum, in its first form. A later cohort was the first to run the 6-week structure that's now documented across the curriculum-week pages. Weeks 1–6 mapped: Foundations / Frame and power / Flight stack and radio / Sensors and first power-up / Flight school / Business and graduation. The structure survived to the current cohort with only minor refinements. The content within each week refined considerably — see below.
The NDVI rig on every build. The Raspberry Pi Zero 2 W rig with red-filtered camera became standard in an earlier cohort. The capture script was developed during the gap between cohorts 01–02; trainees built and integrated their own rigs as part of Week 4. Cooperatives engaged with program graduates responded much more strongly to NDVI false-colour maps than they had to the first cohort's RGB photos: every later-cohort graduate who delivered a first NDVI survey was invited back for a second.
Mission Planner (INAV) introduced. The first cohort used BetaFlight throughout. BetaFlight doesn't have a native Mission Planner tab; trainees attempting autonomous waypoint missions in the first cohort had used custom Python scripts feeding the FC over MAVLink, which was fragile and outside the trainees' skill range. A later cohort introduced the BetaFlight – INAV migration during the weekend between Week 4 and Week 5, gaining INAV's native Mission Planner. The current curriculum still uses this exact approach.
The tether protocol, formalised. Week 4 Friday's tethered first hover became the standard. 1m tether, 30cm altitude, 30 seconds, instructor at LiPo bench within 2m. A later cohort had zero near-miss prop strike events Week 4. The protocol was extended to cover Thursday's first props-on power-up (5% throttle, tethered, no actual flight) — also originally a first-cohort omission.
Saturday make-up sessions, in practice. A later cohort used Saturday sessions across multiple weeks. Week 1: 1 trainee for soldering practice. Week 2: 1 trainee for half-drone diagnostics. Week 3: 2 trainees for BetaFlight Modes-tab work. Week 4: no trainees needed Saturday (the lowest-Saturday-rate week, as documented in curriculum-week-4.html). Week 5: 2 trainees for solo survey mission make-up. Every later-cohort trainee who attended a Saturday session graduated. The supportive philosophy proved out.
The first paid survey — and what went wrong. A later cohort introduced the paid-survey concept in Week 5, not Week 6. The thinking was that doing a paid survey before the business curriculum would make Week 6's contracts and pricing material feel concrete. The execution went badly. The cooperative engaged for the survey expected a higher-quality deliverable than the trainee — flying their first non-practice mission — could produce. The NDVI map had a coverage gap (the trainee's AOI polygon didn't fully match the cooperative's plot boundaries) and an exposure issue (overcast sky during capture). The cooperative manager refused to pay the contracted amount. Lumipad staff intervened, mediated a partial-payment agreement, and the partner relationship was preserved — but the trainee experienced their first "real client" interaction as a conflict, not a success. The lesson was specific: paid surveys belong in Week 6 only, after the trainee has done the Week 5 solo survey assessment over a low-stakes cohort-program plot. The current curriculum reflects this gating.
The 22-item pre-flight checklist, developed mid-cohort. A later cohort had a Week 5 mid-mission radio-loss event during the autonomous waypoint exercise — the failsafe triggered correctly and the drone returned via RTH, no damage, but the team realised that there was no formal pre-flight checklist; what counted as "ready to fly" varied by instructor and by day. Over the weekend after the radio-loss event, the lead instructor and the lead engineer drafted a 22-item checklist drawing on every near-miss and configuration error from the first cohort + the later cohort's incident. The checklist was used for the rest of that cohort and is still the cohort default current checklist (see Week 4 Thursday's drill).
The partner cacao farm relationship. The cooperative manager who offered a later cohort access to their cacao plot is still the cohort program's primary partner. The reciprocal arrangement — cohort fees include a small honorarium plus delivery of all cohort-flown NDVI imagery to the cooperative — has held across three cohorts. The cooperative has used the imagery to make several actual decisions about plot management; the relationship is producing value in both directions.
program graduates-as-recruiters. By the time the team was recruiting for the current cohort (mid-2024), a later cohort had been graduated for ~6 months. Three program graduates who had returned to their home cooperative networks became active recruiters for the current cohort — pitching the program to neighbours, identifying candidates, vouching for the cohort default. The current recruit.html describes graduates-as-recruiters as the highest-yield channel; this came from an earlier cohort graduates.
The graduation ceremony format. A later cohort was the first cohort with a formal graduation ceremony — certificates, drone hand-off, cohort group dinner. The format mostly survived to the current cohort; what changed was scale (the panel grew from 4 reviewers in an earlier cohort to 10 in the current cohort) and the addition of family transport stipends.
Specific changes the current cohort made because of a later cohort
The current cohort is running as this page is written. Most of the curriculum library was written from inside it — we documented what we were doing as we did it, with the structure refined from an earlier cohort and the lessons learned absorbed from both prior cohorts. The 6 detailed week pages, the business reference, the field operations guide, the emergency-response protocols, the parts and combinations references — all reflect the cohort default model. This is the program that future cohort organisers can run with confidence; it has been earned across three cohorts.
Scale to 12 trainees in two cells. First cohort at this size. Instructor capacity is the constraint — the cohort default 1:6 instructor-trainee ratio is preserved, but coordinating two cells in parallel adds operational complexity. The program coordinator role is new in the current cohort; in earlier cohorts the lead instructor handled coordination directly.
The full library documented. earlier cohorts ran with internal notes and instructor knowledge but no public-facing documentation. the current cohort is the first cohort where the curriculum, business model, technical references, and operational procedures are fully written up. Partner organisations evaluating Lumipad have a body of public material to reference; future Lumipad expansion (or analogous programs run by partner cooperatives in other regions) has a starting point.
Two backup partner cooperatives. The primary partner farm is still the same as in a later cohort. Two additional partner cooperatives in adjacent barangays now have arrangements with Lumipad — providing rotation options for cohort field weeks, expanded paid-survey opportunities for Week 6, and additional graduates recruitment networks.
Refined NDVI capture script. A later cohort used a fixed-exposure capture; the current cohort added auto-exposure with bracketing (3 exposures per capture, best one selected post-flight). Reduces capture-quality variability across changing field conditions.
Monthly group call for the current cohort's graduates, first 6 months. In addition to the 30/60/90-day individual checkpoints inherited from an earlier cohort, the current cohort's graduates will have a monthly group video call for the first 6 months post-graduation. Designed to build cohort-to-cohort graduate community connection — program graduates can mentor the next cohort; the next cohort will eventually do the same for future cohorts.
Two completed cohorts isn't enough to know everything. We're writing this section to keep ourselves honest about what hasn't been tested yet. Future cohorts will answer some of these; some are properly Lumipad's strategic questions rather than the cohort program's.
Does the model scale beyond 12 trainees per cohort? the current cohort is testing the 12-trainee size. We don't know if 16 or 20 would still preserve the individual-attention discipline that's been a differentiator. The cell structure helps but instructor capacity is the real constraint; hiring and training new lead instructors takes ~12 months of program participation.
Does the curriculum transfer beyond cacao / coffee / coconut / abaca? The four crops the partner cooperatives grow have specific NDVI signatures that the program has calibrated against. Other Mindanao crops (rice, banana, sugarcane) would require different interpretation; some may need different flight parameters (rice paddies have water reflection issues at certain angles). We have anecdotes from individual graduates who've flown rice surveys but not enough data to write a cohort default reference for it.
Do Year 2+ graduates microenterprise survival rates hold? The 18-month figure (~70% of program graduates operating profitably) is a snapshot at the time of writing. Some graduates businesses have natural challenges in Year 2 — equipment replacement (drone lifecycle ~2-3 years), market saturation in their home barangay, family circumstances changing. The cohort default 30/60/90 cadence covers months 1–3 well but support past 12 months is largely informal.
Is the 30/60/90-day cadence right? Some program graduates want more frequent contact; some find it intrusive; some want it on different topics (technical vs business). The current cohort's monthly group call is a partial response but isn't personalised. We don't yet know what the cohort default support cadence should be.
Does the program work for non-Mindanao regions? Lumipad is a Mindanao-rooted program; partner cooperatives, the partner cacao farm, the workshop, the graduate community are all Mindanao-based. There is interest from cooperatives in Visayas and Northern Luzon; whether the cohort default model transfers — or whether each region needs its own variant — is an open strategic question for Kennemer Foods and Lumipad together.
Is the cost structure sustainable without partner-cooperative honoraria? The current ~₱60–70K per-trainee cost depends on the partner cacao farm relationship being reciprocal (small honorarium + NDVI imagery delivery rather than full commercial rate). If we expand to regions without similar pre-existing partner relationships, per-trainee cost may rise back toward earlier-cohort levels.
What happens after graduates outgrow the program's active support? program graduates at 18 months out are increasingly autonomous; the graduate community is increasingly informal. We don't yet have a model for what "former trainee, now established peer" looks like as a structured relationship. program graduates returning as the next cohort panel members and recruiters is the closest thing; we expect this to become more structured over time.
This page will be updated after each cohort completes. The story above is current to the start of the current cohort; an updated version including the current cohort's lessons will be written after graduation in mid-2026.