One-line verdict. The cohort's most ambitious and fastest-expanding student. Project initiative is now firmly resolved; the open question has shifted to whether his understanding and debugging can keep pace with what he ships.
2.98/5
Competent
Period-average weighted score across all 7 pillars. Scale: 4.5+ Outstanding · 3.5+ Strong · 2.5+ Competent.
How this is scored
This mirrors the program’s evaluation framework (see the site Evaluation section): seven
weighted pillars, each rated 1–5, plus the prompt-quality metrics and L1–L5 maturity levels the
trainers track weekly. Prompt Engineering carries the heaviest weight — in an AI-first program, how you
communicate with AI determines everything else.
Pillar weights
Pillar
Weight
Prompt Engineering
30%
Problem Solving
20%
Code Quality
15%
Security
10%
Accessibility
10%
Professional Skills
10%
Leadership
5%
Score scale
Score
Level
Meaning
5
Exceptional
can teach, innovates
4
Independent
solo, understands why
3
Guided
needs help, gets basics
2
Awareness
knows concept, can't do
1
Absent
nothing
Trajectory at a glance
Prompt quality score
Rises steadily from 20% (W1) to 50% (W7). Graduation target is 60% first-pass-quality.
Prompt maturity level
L1 vague → L5 strategic. Target: L3 by W8. GB reaches L3 within the period.
Security & accessibility in prompts
Both climb late as real backends and the shared app force the issue. Program target is 50%+ by week 12.
Pillar radar — W1 vs W7
The shape expands outward over the seven weeks — the model the trainers overlay monthly.
Weekly prompt tracking
Week
Prompt Score
First-Pass
Security
A11y
Level
W1Sep 1–7
20%
10%
0%
0%
L1
W2Sep 8–14
34%
28%
5%
0%
L2
W3Sep 15–21
36%
25%
5%
0%
L2
W4Sep 22–28
40%
35%
25%
5%
L2
W5Sep 29–Oct 5
46%
40%
20%
5%
L3
W6Oct 6–12
48%
45%
15%
20%
L3
W7Oct 13–19
50%
50%
45%
45%
L3
Period scorecard (average)
Pillar
Score /5
Weight
Weighted
Prompt Engineering
3/5
30%
0.90
Problem Solving
3/5
20%
0.60
Code Quality
3/5
15%
0.45
Security
3/5
10%
0.30
Accessibility
2/5
10%
0.20
Professional Skills
3/5
10%
0.30
Leadership
3/5
5%
0.15
TOTAL
100%
2.90/5
Weighted total above uses rounded period-average pillar scores; the headline 2.98/5 uses the exact averages.
Week 1 · Sep 1–7
Foundations & tooling
Pure environment setup — no web pages yet.
1.75/5
Developing
baseline
20%
Prompt score
10%
First-pass
0%
Security
0%
A11y
L1
Maturity
Pillar scores — W1
Pillar radar — W1
What was built
Prompt-logging script, GitHub repo connection, .gitignore for secrets.
Two assignment-style Kiro steering files (English-only output; a line-by-line code explainer).
Endless push/pull drills to get comfortable with the remote.
Almost entirely mechanical. The real signal is genuine curiosity — a steady stream of “why” questions: difference between HTTPS and SSH, what an upstream repository is, what git identity means.
“by any chance my files and html text file delete from github store. Is there any chance to get it back?”— 09-07
Prompt-rubric read
Prompts are mostly one-liners with no context or constraints (rubric axes 1–4 near zero). The two steering files are structured but are the session-02 assignment template, not his own.
Did well
Asks conceptual “why” questions from day one — the Leadership-pillar “asks why” trait.
Watch / fix
Shaky git mental model; lost work on 09-07 and didn’t understand recovery.
Repeats the same push/pull several times in minutes — doesn’t trust that an action completed.
Week 2 · Sep 8–14
First real builds
His most productive foundational week — first self-initiated app appears.
3.15/5
Competent
▲ improving
34%
Prompt score
28%
First-pass
5%
Security
0%
A11y
L2
Maturity
Pillar scores — W2
Pillar radar — W2
What was built
Personal page + About Me (assigned brief).
Study Notes Maker — his first self-initiated project (notes CRUD, search, favorites, localStorage, an escapeHtml() XSS guard).
The key split: the SHS series is verbatim from the instructor prompt bank (DB-verified) — niche + copy only. But Study Notes Maker and the SHS feature requests (hide-on-scroll, scroll-to-top, dark mode) are genuinely his own.
“This is my first custom project … you will create a very attractive webpage naming it Study note maker.”— 09-10
“if my prompt is not defined don't accept my prompt. Just mention me the prompt is not defined.”— 09-14
Prompt-rubric read
His own SHS feature prompts now carry Context/Task/Format/Edge-case structure (rubric 1,2,5 rising). Security (axis 7) appears once via the XSS guard; accessibility (axis 8) still absent.
Did well
First self-initiated project (Study Notes Maker) with a security guard he asked for.
Explicitly asked to be graded on his prompts and to have under-specified prompts rejected — real metacognition.
Working JS (localStorage, filter/sort/reduce) before JS was taught — self-driven going-beyond.
Watch / fix
Broken nav anchors (#about/#contact point to nothing).
Committed a machine-local image path file:///C:/Users/....
Filename with spaces (Apple 18 pro max.jpg); struggled to give surgical edit instructions.
Week 3 · Sep 15–21
Solo e-commerce marathon
A multi-day solo push building his own e-commerce sites.
2.75/5
Competent
▬ steady
36%
Prompt score
25%
First-pass
5%
Security
0%
A11y
L2
Maturity
Pillar scores — W3
Pillar radar — W3
What was built
EyeStore / “NETRA KUL” eyewear store — home, category & product pages, login/signup modal, cart slide-out, filters, RON pricing, discounts, reviews, fixed social rail, EN/RO toggle.
Started ASS (study-abroad) with per-country/university detail pages.
Skills demonstrated
Ambitious CSS/JS — fixed-position rails, modal/overlay flows, cart animations, product routing, responsive fixes. Clearly past the taught syllabus.
Initiative & origin own work
This decisively answers the first eval’s doubt about thin project initiative. EyeStore and ASS are his own ideas and prompts. He even asked for “more than 100 project ideas I can work as practice” — generating his own briefs.
“make it stylist this is all in your hand”— 09-16
“I can't see image on it. solve this problem.”— 09-16
Prompt-rubric read
Specificity (axis 1) drops as prompts become vibe-led; iteration efficiency (axis 9) is poor — the image bug alone took many “still not working” rounds. Critical assessment (axis 10) weak: debugs by assertion.
Did well
Strong, genuine project ownership — two self-conceived sites in one week.
Actively seeks his own practice briefs (Leadership-pillar initiative).
Watch / fix
Images never work first time — pastes Windows file paths and Google page URLs (not image URLs); a real conceptual gap.
Prompts turn vibe-driven (“make it stylist”, “something crazy”) — leans on the model’s taste.
Git-identity confusion resurfaced; folder-nesting mistakes he had to unwind.
Week 4 · Sep 22–28
Full-stack pivot
Jumps from static front-end to a real full-stack app.
3.05/5
Competent
▲ improving
40%
Prompt score
35%
First-pass
25%
Security
5%
A11y
L2
Maturity
Pillar scores — W4
Pillar radar — W4
What was built
A PHP + MySQL + Angular blog: login with password_verify() + prepared statements, posts CRUD, a JWT guard (public reads, token-gated writes).
Angular login/admin table with edit/delete modals, categories, image upload, search.
A design-system.md steering file for a two-person build (precursor to the Landaro teamwork).
Skills demonstrated
PHP, MySQL schema, prepared statements, password hashing, JWT auth, REST CRUD, Postman, Angular components/services/routing — a massive stack expansion in one week.
Initiative & origin assigned
The PHP/JWT/Angular prompts carry copied coaching annotations (“Why it works: …”), marking them as instructor briefs. His own contribution is the persistence to get a painful Node/PATH toolchain working.
“Update my PHP login so that, when the password is correct, it issues a signed JWT … add a reusable check that protects my other endpoints.”— 09-24
“blog post api is not running check and fix it”— 09-27
Prompt-rubric read
Security (axis 7) jumps thanks to the JWT/hashing brief. But those are assigned prompts, so prompt-quality credit is limited; iteration (axis 9) and critical assessment (axis 10) stay weak.
Did well
Real security practice enters his work: hashing, prepared statements, JWT, “don’t reveal which field was wrong.”
Huge breadth gain — first time touching auth and databases.
Watch / fix
Debugs by assertion (“api is not running check and fix it”).
Dumped ~25 bullets as one run-on prompt twice; a near-miss scaffolding into a teammate’s folder.
Week 5 · Sep 29–Oct 5
Documentation-first discipline
A clear shift to planning before coding — his best process of the whole period.
Landaro / News module planning — 10 sections, a GUID scheme (dropping I/O/0/1 to avoid confusion).
Skills demonstrated
Requirements engineering, information architecture, DB-field planning, challenging his own tech-stack choice.
Initiative & origin own work
High and clearly his own — the docs-first method and “ask me A-to-Z, don’t guess” instruction are self-driven. The single most encouraging trajectory signal in the dataset.
“before we write any line of code, help me plan it out properly … write one part at a time, show it to me and wait until I say ‘fine’ or ‘edit this’.”— 09-30
Prompt-rubric read
Specificity, context and constraints (axes 1–3) are now strong and self-authored. This is the week his prompt-engineering score is most clearly his own.
Reaches L3 maturity: full spec + patterns, deferring payment/SMTP deliberately.
Watch / fix
Typos persist (coading, SMTO for SMTP).
Repaste-spam of the same giant prompt 2–3×; lingering push uncertainty.
Week 6 · Oct 6–12
Light week, one strong prompt
Low volume, but the one substantive task is his cleanest independent prompt.
3.20/5
Competent
▬ steady
48%
Prompt score
45%
First-pass
15%
Security
20%
A11y
L3
Maturity
Pillar scores — W6
Pillar radar — W6
What was built
Mostly pushes + a workflow automation (“tell me ‘pushed is done to git’ … remember this forever”).
Made ASS mobile-first — base phone styles, min-width breakpoints at 768/1024px, flexible units.
Skills demonstrated
Mobile-first responsive refactor with named breakpoints and a review checklist.
Initiative & origin own work
The 10-12 mobile-first prompt is textbook and self-authored — the prompt-quality lessons from W2 and W5 have stuck.
“Rework this page to be mobile-first … use rem/%/fr/clamp instead of fixed pixels … list exactly what you adjusted and why, so I can review it.”— 10-12
Prompt-rubric read
Output format and iteration-efficiency axes (5, 9) are excellent in the one prompt that matters. The concern is consistency, not ceiling.
Did well
His single best independent prompt: specific breakpoints, named units, no-behaviour-change constraint, a request for a reviewable change-list.
Accessibility awareness nudges up (responsive = usable on more devices).
Watch / fix
Very low output volume this week.
Security barely present when not handed a security brief.
Week 7 · Oct 13–19
Real team contribution to Landaro
Contributing the News + Accounts modules to the shared Angular + PHP app.
3.50/5
Strong
▲ improving
50%
Prompt score
50%
First-pass
45%
Security
45%
A11y
L3
Maturity
Pillar scores — W7
Pillar radar — W7
What was built
A 50+ article seed and a live debounced search with an accessible combobox/listbox.
Phone-number verification (Twilio with a disk fallback, an additive migration, EN/FR i18n).
An admin users dashboard redesign (sortable server-side columns, status tabs, stat cards), on feature branches to avoid conflicts.
Honest flag: the W7 logs show heavy use of an orchestrated agent workflow — several “prompts” are machine-generated implementation briefs, not GB typing. Genuinely his: the product direction, the emphatic data-safety insistence, and team-coordination instincts.
“I dont need any collapse and conflict between my mates … keep it clean and unconflict way.”— 10-17
“please create the branch for account and admin … which will help not to make conflict with my mates.”— 10-17
Prompt-rubric read
Security and accessibility metrics finally climb — but largely because the agent pipeline enforced them, not because he authored them. Credit the product direction and git discipline; verify the code understanding separately.
Did well
Professional instincts: branch-per-feature, merge awareness, emphatic shared-DB safety rules.
Security and accessibility now show up as real bars (the workflow enforced a WCAG/a11y quality gate).
Working inside the most professional environment he’s touched.
Watch / fix
The sophistication of the output now substantially exceeds what he can demonstrably write himself.
The old “explain every line of your code unaided” warning applies more strongly to this agent-generated full-stack code.
DevSchool · Summary
Gyanendra Bohora — overall
Seven-week trajectory, strengths, and priorities for the next period
Week-by-week summary
Week
Focus
Weighted
Grade
Level
Origin
W1
Foundations & tooling
1.75
Developing
L1
assigned
W2
First real builds
3.15
Competent
L2
mixed
W3
Solo e-commerce marathon
2.75
Competent
L2
own work
W4
Full-stack pivot
3.05
Competent
L2
assigned
W5
Documentation-first discipline
3.45
Competent
L3
own work
W6
Light week, one strong prompt
3.20
Competent
L3
own work
W7
Real team contribution to Landaro
3.50
Strong
L3
mixed
Strengths & priorities
Clear strengths
Curiosity & the “why” habit — consistent from W1 to W7, including challenging his own tech stack.
Prompt metacognition — asked to be graded (W2), ran docs-first (W5), wrote a textbook mobile-first prompt (W6).
Project initiative — resolved — EyeStore, ASS, Goby and the Landaro planning are all his own ideas.
Breadth — static HTML/CSS → JS UI → PHP/MySQL/JWT → Angular → a shared production-style app in seven weeks.
Priorities for next period
Understanding must catch up to output. He debugs by assertion, not diagnosis. Have him read, run and explain the Landaro Angular + PHP code line by line, unaided.
Fix the chronic asset/path blind spot — Windows paths and page-URLs vs. real image URLs; the committed file:///C:/....
Craft & verification — broken nav anchors (flagged W2, never confirmed fixed), filename-with-spaces, folder-nesting slips. Test every link and asset.
Prompt hygiene — stop the giant run-on repastes and duplicate sends.
Git confidence — understand state instead of wanting obstacles to disappear.
Bottom line. GB is the most ambitious, fastest-moving student in the cohort, and the initiative question from the first evaluation is now closed. The next milestone is not more breadth — it is depth: proving he can read, debug and explain the full-stack code he is now shipping, and clearing the recurring asset-path and craft bugs that have followed him since Week 2.
own work self-authored idea/prompt
mixed partly assigned, partly own
assigned executing an instructor brief