It makes a case from DreamSpace Academy’s own record, states the counter-argument at full strength, and marks how well each claim is evidenced. Where it describes what we run, the core pages are the standing description and this is the reasoning behind them.
An ecosystem is built at the age at which people are still deciding what is possible
One of three essays on how the education and the organisation fit together.
What this is. The case for why an organisation trying to build an innovation and entrepreneurship ecosystem reaches down into Grade 6 rather than starting at the graduate — written from DreamSpace Academy’s own record, including the parts that argue against it.
How to read the evidence marks. Events come from DreamSpace Academy’s own delivery record; decisions from its written model. Grades: 🟢 primary or corroborated across sources · 🟡 partial or single-source · 🔴 asserted, and under validation.
⛔ Read §7 before quoting anything from §2. The developmental evidence for starting early is real and it is 🟡, deliberately — it has published counterweights, they are recorded here, and the claim is “the window argues for early”, never “early is proven.”
1. The organisation changed its mind, and wrote down why
For its first seven years DreamSpace Academy aimed at people aged 18 and over. Bootcamps, incubation, labs, the venture pathway — all of it pitched at young adults who had finished school.
Then it stopped, and restructured. The sentence in DreamSpace Academy’s own concept report is short:
“Intervening at this age proved late for long-lasting transformation.” 🟢
What came out of that finding was a split into Empowerment Inside School and Empowerment Outside School, and EIS — the in-school half — was designed as something the organisation had never attempted before: a sustained six-year engagement, Grade 6 through Grade 11, delivered free into underserved government schools, on the school’s own timetable, using challenge-based learning rather than rote instruction 🟢.
That is not an expansion of the target audience. It is a statement that the previous target audience was the wrong end of the problem, and it is the honest starting point for this article: the claim did not come from a theory. It came from DreamSpace watching its own results and concluding it was arriving too late.
2. What the evidence says about the window
The literature DreamSpace found afterwards agreed with the finding — and this ordering is worth noting, because it means the evidence was not selected to justify a decision already taken.
Entrepreneurial capability is developmentally banded, and the band opens earlier than most programmes start.
- Abstract business concepts are cognitively gated: children cannot reliably separate capital ownership from labour until roughly ages 12–13 🟢; profit understanding is middle-childhood but is instruction-gated on a price-comparison prerequisite 🟢.
- Saving and selling are accessible from around age six in a concrete, playful setting 🟢.
- The most on-point randomised trial — BizWorld, ~2,751 children aged 11–12 — shows robust gains on non-cognitive entrepreneurial skills: persistence, creativity, forward-looking behaviour 🟢.
- The one clean early-versus-late comparison in the field is the decisive one: a randomised field experiment in the final grade of primary school found the same mindset gains, and recorded that the same programme delivered to adolescents produced zero or negative effects (Huber, Sloof & Van Praag 2014) ⚠️ — a lead taken from a literature scan, and not yet in our verified reference corpus.
So the mechanism being claimed is narrow and specific. It is not “children learn faster”. It is that the thing an ecosystem needs — a disposition to try, to persist, to treat a problem as addressable — is movable at eleven and much harder to move at twenty-two. DreamSpace’s four-stage Maker Journey is built on exactly that: Explorer, Maker and Innovator are stages of purpose before they are stages of skill.
3. The system reason: school is the only place that reaches everyone
The developmental argument says early. A second, entirely separate line of evidence says in school, and it is the stronger of the two because it is about access rather than psychology.
Where the barrier is cost and distance, a walk-in makerspace does not reach the child. In Nuwara Eliya, 26.7% of children who left school did so because the household could not afford transport or learning materials, and 13.3% had no means of getting to school at all 🟢. In Mullaitivu the barrier is a different one — 33.3% had not passed O/L, and 11.1% had to work for the family 🟢.
Neither of those children ever arrives at a makerspace. Both are in a classroom until the day they are not. The school is the only delivery point that reaches everyone, including the people least able to opt in — which is why the in-school half of the restructure is the half that carries the inclusion claim.
And the state has just created the opening. Sri Lanka’s curriculum reform makes ICT and “Entrepreneurship & Financial Literacy” compulsory essential subjects in Grades 6–9, phasing in from 2026 and reaching the full band by 2029 🟢 — while only 19% of type-2/3 schools have IT labs against 72% of top-tier schools, and the technology-stream student/teacher ratio is 35.3 against a 12.8 average 🟢. In the estate sector 1.5% of households own a computer and 92.6% of internet access is by smartphone alone 🟢.
A mandate the delivery layer cannot yet meet is a demand signal, not a competitor. It creates a partner-and-feeder position for an organisation that already has trainers, curriculum and lab equipment in those districts.
⚠️ The honest limit on that: whether the position is contractible — an MoU, a provincial agreement, a recognised delivery-partner status — is an open gap. What exists today is real and narrower: written Zonal Education Office approvals in three zones (Batticaloa, Kantale, Nuwara Eliya), and a school-screening rule that deliberately selects for absence — no ICT teacher, no computer lab, low O/L ICT results, rural or marginalised 🟢.
4. The readiness reason: you cannot bolt an advanced module onto an unmeasured baseline
The third argument is the one DreamSpace paid for in delivery, and it is the least comfortable.
Across many sessions, an experienced DreamSpace Academy trainer recorded that the binding constraint on delivery is not equipment 🟢:
“The biggest issue is not the lack of facilities, laptops, internet, or tools… The biggest issue is the mismatch between the expected level of the session and the actual readiness of the participants.”
“In many sessions, less than 10% of the students were actually fit for the planned content.”
A separate diagnostic on interview-selected students found that of roughly fourteen, five could reproduce a demonstrated Scratch password-checker and none built one independently — and the four fastest completers could not explain why the skills mattered 🟢. Completion is not transfer. Attendance decays at the complexity onset.
Set that beside the regional data — estate-sector literacy at 81.3% and falling, and a cohort that is 92.6% smartphone-only — and the conclusion is structural rather than pessimistic:
A learner who has never used a desktop operating system is not a learner who “just needs the advanced module.” Adding more provision on top of an unmeasured baseline reproduces the same decay. The answer is a measured intake baseline, fundamentals-first sequencing, and dose.
Every one of those three is easier to deliver to a child than to an adult, because a six-year in-school engagement has the baseline, the sequence and the dose built into it. A four-week adult bootcamp has none of them, and the evidence on dose is explicit: one workshop produced no significant gains where a multi-session programme did 🟢.
⭐ This is the cheapest high-value fix in DreamSpace’s whole analysis — measure readiness behaviourally, at intake, instead of asking people whether they know the topic. It costs a pre-task.
5. The ecosystem reason: children are the ecosystem’s future staff, and the record shows it
The first three arguments are about learning. This one is about supply, and it is the argument that connects this article to the first in the series.
DreamSpace’s hardest constraint was never labs, funding or curriculum — it was identifying changemakers, because the model waited for already-motivated individuals to surface. The fix was to build motivation through the Maker Journey rather than screen for people who arrive with it, which converts discovery from a search problem into a delivery problem. A school programme running Grade 6 to Grade 11 across dozens of schools is what a delivery answer to that problem physically looks like.
And the return on the earlier cohorts is already visible in the organisation itself:
- Alumni now run the labs, and two are directors 🟢.
- When the model went to Erode, Tamil Nadu, it was “delivered by two people who are themselves products of the pathway” 🟢.
- The reach runs upward as well as downward: DreamSpace Academy’s school work has extended from students to teachers, then to principals and zonal education staff, and most recently toward preschool teachers through a partner that funds preschools in Sri Lanka 🟢.
⚠️ Two limits on that, and both are load-bearing. First, staff running the labs is Batticaloa’s bench, and Erode is that bench travelling — it is not evidence that a new site can grow its own. Second, the one test of the second claim failed: DreamSpace Mullaitivu closed, and the recorded cause was a custodian-succession failure — the person running it could not continue, no local successor with the interest was found, and the search ran in Mullaitivu only. Not funding, not operations, not capability.
⭐ That closure is the sharpest argument in this article, read the right way round. A model whose continuity depends on one custodian is a model that has to be growing its next custodians continuously — which is a statement about how early you have to start.
6. What “starting with children” does not mean
This is where a reasonable-sounding claim usually goes wrong, so the fences are stated explicitly.
⛔ It does not mean the makerspace is for children. Age is not used in DreamSpace’s model at all — not as an entry rule, not as a band, and not descriptively in marketing either, because a typical age printed on a brochure is read as a rule by exactly the people the makerspace is trying not to turn away. Maker education is for everyone. Placement is by evidence, through a placement check.
⛔ It does not mean entrepreneurship is taught to eight-year-olds. Entrepreneurship content is gated by prerequisite, not by age. Early stages teach the entrepreneurial mindset through making plus simple sell-and-save loops; profit-versus-revenue, pricing and venture-building attach to later stages, whenever a person reaches them.
⛔ It does not mean everyone is aimed at founding a company. Changemaker is a fork, not a funnel — impact venture, research, activism or community leadership, employment or freelancing, Grow Your Own, and others are four equally legitimate exits. The evidence base warns specifically that pushing marginal students into venturing can lower their outcomes 🟡, and the model is built to that warning rather than against it.
⛔ It does not mean the school programme is a device donation. A laptop in a school with no ICT teacher does not get used. What is delivered is the machine plus a trained facilitator, a curriculum aligned to the national syllabus, zonal approval, and baseline-to-final measurement.
⛔ And it is not charity delivered into a school. The clearest failure in the record is exactly that misreading: EIS was terminated at Dickoya Tamil Maha Vidyalayam in 2026, and the progress report says why in the school’s own terms — “They see us as an NGO and want us to buy things for the school” 🟢. When the host institution’s expectation is equipment rather than capability, delivery collapses regardless of student demand. In a region used to hand-outs, “free programme” is read as “free things” — which is a design problem for anyone attempting this, not a complaint about one school.
7. The counterweights, recorded on purpose
The early-intervention case is graded 🟡 and must stay 🟡. Two published findings pull against it and are kept in the model deliberately:
- Non-cognitive-skill effects on later outcomes are heterogeneous. A screen of 9,553 publications (Smithers et al. 2018, Nature Human Behaviour) found prediction intervals spanning negative, null and positive, with small-study and publication bias 🟡.
- Early gains can fade. An intensive Rwandan secondary reform raised entrepreneurship at one year but the gain faded by three, with employment 6 percentage points lower — and the losses were concentrated among marginal students pushed into venturing 🟢.
The honest form of the claim is therefore:
The developmental window argues for starting early, and the one clean early-versus-late comparison in entrepreneurship education favours early. That is not the same as “early intervention is proven.”
Three further gaps belong in the same paragraph, because they are DreamSpace’s own:
- No Sri Lanka–specific early-age entrepreneurship impact evidence exists — only policy intent. The best low- and middle-income-country evidence is Rwandan, and its transfer to post-conflict, multilingual, rural Sri Lanka is untested 🟢.
- Sri Lanka does not appear in the maker-education literature at all, across five independent corpus reviews 🟢.
- No clean rural-makerspace-to-income result exists anywhere, which is why the Earn claim is a hypothesis under validation rather than a finding 🔴.
⭐ That last cluster is not only a limitation — it is the strongest thing DreamSpace can offer a research funder. An unmeasured region is an unfunded region. The organisation has an eight-year longitudinal field site, a research engine, a case-study and pattern register, two peer-reviewed papers and a hosted PhD study. Almost no peer can generate this evidence, and DreamSpace Academy can.
8. What the school work has actually delivered
✏️ Updated 2026-08-13 against the
EIS-26re-mine of the EIS archive. Two figures are now available for the same programme and they are different documents answering different questions. ⛔ They must never be merged, and neither may be used as the other.Enrolment = who was signed up. Reach = who turned up.
8a · Reach — who turned up
Quoted from narrative reports only, never spreadsheet row counts, which this archive inflates to 977/999/1000+ through pre-formatted empty ranges:
| Cohort | Period | Reach (verbatim) |
|---|---|---|
| Batticaloa ICT pilot — final impact report | Jul–Dec 2025 | “Reached 194 students across Grades 6–9”, 5 schools |
| Hatton — robotics + ICT | late 2025 | “No of Students: 306” (Dickoya 167 + Ireby 139); paused by floods, resumed 8 Jan 2026 |
| Nuwara Eliya robotics — 4 schools | 23 Apr – 29 May 2026 | “Total Students Reached 114” (90 M, 24 F); 13 sessions; 21 projects |
| Batticaloa Grade 8 — one-month trainer report | 23 Jan – 13 Feb 2026 | “Total Partipates 72” across 5 schools ⛔ misfiled under Hatton in the archive |
| Kantale ICT — monthly training report | June 2026 | 5 schools × 4 h/week; per-session headcounts 85 · 110 · 16 · 55 · 55 (weeks 1/3/4; week 2 blank) |
8b · Enrolment — who was signed up
The EIS-26 pass surfaced the first single document in the repo that counts the whole programme at
once: the 2026 enrolment register, three sheets, one per hub 🟢.
| Hub | Schools | Enrolled |
|---|---|---|
| Batticaloa | 7 | 272 |
| Kantale | 5 | 419 |
| Nuwara Eliya (“DreamSpace Academy Hatton” sheet) — robotics only | 4 | 118 |
| Register total | 16 | 809 |
⛔⛔ 809 is not the 2026 programme size and must not be published as one. The register’s Hatton sheet counts robotics only, and Nuwara Eliya’s two ICT schools are absent from it entirely — St John del Rey T.V (39, per its own June report) and Annfield T.M.V (35, per its 24 Jun 2026 school letter). EIS runs in 18 schools in 2026; the programme’s own register counts 16. A corrected register is an open ask.
⚠️ And Kantale has three headcounts, from three documents in the same folder — school agreement letters 523, register 419, June delivery record 322. Four of the five schools agree letter-to-register and shed ~25–30% by delivery, which is ordinary attendance attrition. Aysha Balika is the outlier and moves at both steps — 253 → 149 → 110 — and whether its 253 was the school’s whole Grade 6–9 roll rather than an enrolled group is unanswered. ⛔ Until it is, 253 is an ask, not a count.
8c · Where it runs
Batticaloa — 5 pilot schools 2025, 2 more added 2026 at the schools’ own request (7 on the register) · Hatton / Nuwara Eliya — Dickoya and Ireby 2025; six schools in 2026, four robotics and two ICT · Kantale — 5 schools, 2026 · Mullaitivu — planned, not implemented, still awaiting zonal approval 🟢.
8d · ⚠️ The finding that argues against this article’s own thesis
The EIS-26 pass surfaced something an article titled “Why It Starts With Children” has an
obligation to print: in 2026 the target band drifted upward, and DreamSpace Academy’s own trainer recorded it in
writing. Of the six Nuwara Eliya schools, Sri Pada College ran Grades 10, 11 and 12 and
Fruithill runs to Grade 11 — against a stated Grade 6–9 target. The trainer’s note is blunt:
“The school was giving us a set of students who are not from our targeted grades… the students are adults… the delivery language is sinhala.”
The school’s own explanation is capacity — it “can not give only selected students as their grade 6 to 9 classes have a huge amount of students.” And DreamSpace Academy wrote its own corrective: “Schools should be clearly informed about the target student profile (Grades 6–9) before the programme starts.” 🟢
⭐ Read it as what it is. The organisation concluded that starting at eighteen was late, designed a Grade 6–11 programme in response, and then — at two schools in one hub in one term — delivered to adults in a language its trainer does not teach in. The design intent and the delivered reality came apart at the point of school scheduling, which is exactly where a partner-delivered programme is weakest. ⛔ It is not evidence the thesis is wrong. It is evidence the thesis is not self-enforcing, and that the fence has to be written into the school agreement rather than assumed.
8e · Cautions on every number above
- The concept report’s “616 students across 9 schools in 3 regions” is a planning target that includes unimplemented Mullaitivu — a plan, not delivered reach.
- The Story Deck’s headline school figures (50+ schools · 5+ zonal offices · 2,000+ students · 300+ teachers) are self-reported and recorded as “DreamSpace Academy states”, never as verified counts.
- Cohort totals have never been de-duplicated, so no aggregate across programmes is defensible.
- ⛔ The archive’s folder tree is not trustworthy about which hub, month or grade a file belongs to. Six filing errors were found in one pass, four of which produce a wrong number rather than a wrong label — a Batticaloa report filed under Hatton, a Nuwara Eliya report duplicated into Batticaloa, folder month disagreeing with filename month and content month, and Excel converting grade ranges into date serials. Open the file and read its header; never trust its path.
- ⛔ Mid-term and final assessments are marked Mandatory in DreamSpace Academy’s own document checklist and do not exist in the 2026 folder. The baseline-to-final measurement described in §3 is the design; for 2026 the second half of it is currently missing.
Who funds it matters to the argument too. EIS is free to schools and co-funded by at least three micro-donors, each contributing partially: IFNB (Germany) at €6,500 for the 2025 first round, Micro-Tech CNC Pvt Ltd (India) at INR 700,000 under a signed agreement running Sep 2025 – Aug 2026, plus a Sabrina Yousuf Foundation micro-grant (LKR 300,000 disbursed, 27 Nov 2025) 🟢. The programme’s own stated sustainability requirement is “minimum 5,200 EUR / 2M LKR monthly” per hub, for roughly 15 staff, infrastructure and materials — a stated figure, recorded as stated, not analysed here.
9. The claim in one paragraph
DreamSpace Academy did not decide to work with children because children are appealing to funders. It decided it after concluding, from its own results, that starting at eighteen was late — and the developmental literature it found afterwards agreed, with caveats it has kept rather than dropped. Three further reasons stack on top of the first: the school is the only place that reaches the child whose barrier is transport money; readiness, not equipment, is the binding constraint on every programme DreamSpace Academy runs, and only a multi-year in-school engagement has the baseline, sequence and dose to fix it; and an ecosystem that cannot grow its own custodians closes when one person leaves, which DreamSpace has already watched happen once. The children in a Grade 6 ICT class in Kantale are not the ecosystem’s beneficiaries. On this model’s own logic, they are its future staff — and the organisation is honest that it cannot yet prove the rate at which that conversion happens.
A note on sources
Delivery evidence comes from DreamSpace Academy’s own record of the in-school programme — the concept report, the thirteen-phase standard operating procedure, locations, reach, funders and staffing — together with the 2026 re-examination that added the enrolment register, the hub-by-hub delivery record, the document checklist and the filing-error register. Also the funder history, and the case studies on the terminated estate-sector site and on learner readiness. The model material is the root-cause analysis, the identity specification, the DreamSpace Lifecycle, the maker levels and the curriculum, and the recorded decision that removed age from the model.
Statistics resolve to their publishers: the Department of Census and Statistics Labour Force Survey 2023 and Computer Literacy Bulletin 2024, UNDP and OPHI district profiles 2023, and the World Bank Sri Lanka Development Update (October 2024). Intervention evidence is cited to the published studies named in the text; the full list is on the references page.
⚠️ Some figures circulating near this material are refuted, and none of them appears here.
⛔ Two framings this article deliberately avoids, and which must not be reintroduced. The North and East do not have Sri Lanka’s worst youth unemployment — the Northern Province has the lowest, at 14.9%; the real constraints are underemployment and informality in the East, and entrapment in the estate belt. And the multidimensional vulnerability index (2022–23) and the multidimensional poverty index (2019) are different instruments, and must never be presented as a time series.