00 / origin
Dotportion

Dotportion is a coding assessment where candidates build real features in VS Code — with an AI agent, on a token budget.

Whiteboards. LeetCode. Take-homes.
leetcode-style-interview.example
// reverse a binary tree — 45:00 remaining

Traditional interviewsmeasure yesterday’sengineers.

Algorithm puzzles can’t tell you how someone ships production code with modern tools.

Their editor. Their AI. A real codebase.
src/parser.tssrc/index.tstest/parser.test.tsREADME.md
1// parse a CSV line, quotes and all
2export function parseLine(line) {
3  const out = [], let field = "";
4  let quoted = false;
5  for (const ch of line) {
6    if (ch === '"') quoted = !quoted;
7    else if (ch === ',' && !quoted)
8      out.push(field), field = "";
9    else field += ch;
10  }
11  return out.push(field), out;
12}
Dotportion AI
Handle quoted commas inside CSV fields.
Added a state machine for quoted fields + a failing case to the tests.
token budget40%

Modern engineersbuild with AI.

Their own editor. A built-in AI agent. A metered token budget. Exactly like the job.

Prompts · edits · pastes · terminals · test runs
capture · attempt #4821 · live
00:02prompt“Handle quoted commas inside CSV fields.”
00:03aipatch proposed · src/parser.ts
00:04editsrc/parser.ts · +14 −2
00:05paste42 lines → matched to external source
00:07terminalpnpm test
00:07test ✗3 passed · 1 failed
00:09editsrc/parser.ts · +3 −1
00:10test ✓4 passed · 0 failed

Every decisiontells a story.

Every prompt, edit, paste, terminal command, and test run — captured as evidence.

Four categories. Twenty-two dimensions.
0
0.30 · 6 dims
Judgment
0.35 · 7 dims
Output
0.20 · 5 dims
Process
0.15 · 4 dims
Efficiency

Not vibes.Evidence.

22 measured dimensions across Judgment, Output, Process, and Efficiency.

Deterministic scorers + versioned LLM judges
prompt_craft
78
trust_calibration
91
delegation_balance
77
production_mindset
90
requirements_capture
76
tradeoff_reasoning
89
harness_correctness
78
code_quality
91
architecture
77
error_handling
90
documentation
76
naming_clarity
89
idiom_fit
75
verification_discipline
78
debugging_adaptation
91
paste_discipline
77
testing_discipline
90
time_allocation
76
budget_efficiency
78
ai_burst_pattern
91
iteration_velocity
77
redundancy_avoidance
90
0 / 100
Strong hire
Judgment · w 0.3088
Output · w 0.3594
Process · w 0.2083
Efficiency · w 0.1590

One defensiblescore.

A deterministic engine turns the evidence into one defensible, citable score.

Ranked candidates. Cited reports.
candidates
47
avg score
81
top score
94
anomaly rate
6%
01A. Okonkwo2 anomalies94
02M. Larssonclean run91
03R. Venkatclean run88
04S. Haddad1 anomaly84
05J. Parkclean run79

Built for hiringteams.

Ranked candidates. Cited reports. Anomaly flags your whole panel can trust.

White-label, down to the pixel.
Dotportionassessments · backend cohort
candidates47
avg score81
pass bar75
A. Okonkwo94
M. Larsson91
Invite candidates →
DotportionNorthstarVerdantEmberOrchid

Make ityours.

Candidates see your brand, not ours — your logo, your colors, your assessment.

Now onboarding ten design-partner teams

AI makes every candidate look good.We show you who actually is.

Join the waitlist →
DotportionCoding assessment for the AI era

09 / join

Be one of the first ten.

We’re onboarding ten design partners ahead of public launch. Free during the build period, $99/month for six months after (50% off our planned price). Tell us about your team and we’ll be in touch.

01

Real work

Candidates build a real feature in VS Code, with an AI agent, on a server-enforced token budget.

02

Total capture

Every prompt, edit, paste, terminal command, and test run is recorded as evidence.

03

Defensible score

22 dimensions across 4 categories roll up into one citable report your whole panel can trust.

Waitlist signup

We read every response personally. We’ll reply within 48 hours if you’re a fit for the design partner cohort.

Dotportion is built by Sri Charan, a solo founder. Every waitlist submission is read personally — no drip campaigns, no SDRs. If you’d rather talk first, book a 20-minute call →

Dotportion · Coding assessment for the AI era
skip → waitlist