dev/social · product pitch · Alexia Moreanu

AI made code cheap.
It made trust expensive.

Merged pull requests on GitHub, per month

The problem

Maintainers are closing the door to strangers.
Every newcomer is a stranger.

My bet

dev/social: a social layer before the pull request. Show your work in 60s, find people through people, knock before you push.

Why Google

Microsoft owns GitHub and LinkedIn. Google has no home for developers, but it does have YouTube Shorts, Gemini, and 1,000+ developer groups.

VISION Every builder is one warm intro away from the right project.

GitHub PR volume: GitHub via CodeRabbit (2026); 2025 average: Octoverse 2025; PR controls: The Register. Prototype: github.com/alexia-moreanu/dev-social

1 / 5

02 · User & pain

Building alone isn’t a choice. It’s the default.

L

Lena, 23

Junior dev · the solo builder

3 side projects 0 collaborators wants in on OSS

“I’d love to contribute, but I don’t know anyone there. I don’t want to be the spam PR.”

Other side: Sam, maintainer

60%of maintainers are unpaid ~60%have quit or considered quitting 2 in 3popular projects rest on 1–2 people

Sam needs help. He just can’t vet strangers.

Lena’s journey today: where it leaks

71.6%

of GitHub repos are personal, one-person projects

230/min

new repos. Discovery is a Hacker News lottery or a Google search

7 : 10

tried-and-rejected for every 10 accepted contributors. Top reasons: duplicate work, vision mismatch

½

of contributors contribute once. 30% say a rejection stopped them

People join people.

Developers join projects where they already know someone. Social ties plus the right skills mean +30–54% productivity. Missing context, not missing code, is why newcomers bounce.

NOT IN V1 → Big foundation projects (already have GSoC/LFX) Recruiters (turn builders into candidates) Non-developers

Kalliamvakou et al., MSR 2014 · Octoverse 2025 · Steinmacher et al., ICSE 2018 (10,099 quasi-contributors, 21 projects) · arXiv 2026 · Casalnuovo et al., FSE 2015 (+29.5–54.3%) · Tidelift 2024 · Avelino et al., truck factor, 133 projects. Lena and Sam are composite personas.

2 / 5

03 · Market & strategy

Nobody owns the corner where proof of work meets people.

Positions are my judgment. User counts are public figures.

Why Google: it’s a gap in Google’s map too

Trends working for me

+36M

new developers on GitHub in one year

1 in 3

new GitHub devs from India by 2030, mostly mobile-first

46%

of devs distrust AI accuracy, so human trust carries a premium

70%

of people learning to code learn on YouTube: video is how devs learn

The risk I take seriously

Google+ had 90% of sessions under 5 seconds. Lesson: don’t build “social for everyone.” Build one job, for one group: finding people to build with. Polywork’s shutdown says the same thing: a nicer profile isn’t a reason to return.

Octoverse 2025 · LinkedIn stats · Kaggle Chronicles 2025 · YouTube Shorts · GDG · Stack Overflow Survey 2025 · Google+ shutdown · Polywork

3 / 5

04 · Solution & MVP

One loop: Show → Find → Knock → Build → Vouch

SHOW · 60s clip

▶0:42

kvlite · Rust KV store
⎇ mirak/kvlite · ★ 342

Looking for: 1 dev who knows compaction

FIND · your web

Warm path: you → Ada → Sam

KNOCK · not a cold PR

to Sam · via Ada
“I use kvlite at work. I’d like to take issue #42 (tombstone GC).”

Let inNot now

Sam’s inbox: 3 / 5 knocks today · cap set by Sam

After first merge → Sam vouches for Lena

Priorities (RICE, my estimates)

Project + 60s clip

9.6 Knock (capped)

8.4 Web · warm paths

5.6 Vouch after merge

4.8 Gemini starter-task match

3.8 Tips feed

2.9

MVP next

Cut on purpose

Hosting code — my original idea. GitHub does it well and nobody migrates repos. Meeting is the gap.

News page — I built it. HN and daily.dev own it.

Recruiters — builders who feel like candidates start performing.

Working prototype (Next.js + Prisma) runs the full loop: 60s clips, the Web graph, Knock with daily caps and warm paths, and Vouch. Next: verifying merges on GitHub. RICE = Reach × Impact × Confidence ÷ Effort; scores are my estimates. Graph view inspired by Obsidian.

4 / 5

05 · Go-to-market, metrics & risks

Start with one dense network. Measure collaborations, not likes.

North Star

New collaborations / week

knock accepted → merged PR in 30 days, verified on GitHub

Projects posted with a clip

Knocks sent per active builder

Knock → accepted rate

Accepted → first merge rate

Counter-metrics I watch

🧯 Knocks per maintainer / week, mute rate

🤖 Share of knocks flagged low-effort / AI

📈 % of knocks going to the top 1% of projects

👀 Clip views with no repo action (vanity)

Risks

1 Cold start. Campus networks where one persona plays both sides.

2 Slop moves to knocks. Maintainer caps, context required, vouch-weighted ranking. In MVP.

3 GitHub copies it. Distribution via Shorts + Android, and the graph compounds.

4 Google’s social track record. Narrow job, GitHub-first, open data export.

5 Performative content. Every clip must link to real commits.

Help me validate the rest. Developer or maintainer? Tell me where this breaks. Give feedback →

Phase gates, targets and risk placement are my proposals, not measured results. Baseline: “half contribute only once” (arXiv 2026). Feedback form is a public GitHub issue form, so please don’t include personal contact details.

5 / 5

AI made code cheap.It made trust expensive.