Mo Suvo (@MoSuvo) header
Mo Suvo (@MoSuvo) profile picture

// what broke, what shipped, what nobody tells you // building SaaS + AI · design roots, systems underneath ↳ peekover: open source markdown reader · dad of two

Joined November 2020
Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)Photo by Mo Suvo (@MoSuvo)
Pinned Tweet
I built a markdown reader that updates live while your agent writes. peekover, now on Chrome and Firefox: - see exactly what the agent changed - tables, diagrams, math - zero browser permissions Free and open source. v0.1, so tell me what breaks. #buildinpublic #opensource
11
9
286
Hey @X, let's connect. Reply with two things: 1. what you're building 2. what you do when you're not building I'll go first: 1. peekover, a free markdown reader I shipped yesterday 2. dad of two, so mostly taking care of them. I'll check out every reply. #buildinpublic #ai #connect
24
18
473
I am not switching to openai any time soon!
I will not be writing a full review of GPT-6.2 Astra / Sol. I had a terrible experience testing them. They were essentially unusable. I do not recommend these models.
1
57
Describe your current to-do list in one word. Mine: "optimistic."
2
4
55
AI will happily run your tests all day. It'll never tell you they're testing the wrong thing.
1
3
67
Your first version: A) Small and polished. Does one thing well. B) Rough and does everything. Fix it later. Which one did you actually ship?
4
4
177
Cancel and Submit look exactly the same. One of them takes $248 from you. Which one? That's how I can tell an MVP was AI-built. Three small tells, ten minutes to fix. #UXDesign #vibecoding
2
6
151
Designer habit I never lost: squinting at the screen. Whatever still stands out when it's blurry is what people see first. #UXDesign
1
3
75
GitHub has a setting called "Prevent self-review". I had to leave it off. When you're a solo dev, the only reviewer is you. #buildinpublic
11
1
8
221
Useful data. The finding I’m keeping: “opening with something you built or tested” beats almost everything else. Good timing, since I’m about to launch something.
How-to posts are 3× more likely to break out on X. > I analysed 659 posts from 10 growth accounts to find out what gets reach. I measured each post against its author’s usual views. Anything above 5× counted as a hit. Only 62 posts made it. A 9% hit rate. Here’s what performed better: • How-to posts: 30% hit rate • Opening with something you built or tested: 27% • Videos: 26% (vs. 6–9% for other formats) • Tagging someone: 17% • Job announcements: 5 out of 10 broke out Each pattern worked across at least 3 accounts. No single creator carried the results. What never helps: • Starting with a number or dollar amount • Opening with a question Articles looked promising: 6 hits in 7 posts. But all six came from one account. Too little evidence to call it a pattern. One more finding: 97% of breakout posts were reposted or quoted at least once. My takeaway: stop posting what you know. Show people exactly how you did it.
2
5
108
Things Firefox's add-on store taught me this week: 1. The summary has a 250 character limit. Mine is 247. 2. You upload your full source code with it. 3. You can't rate your own add-on. 4. "Submit a New Add-on" fails if you've submitted before. You want "Upload New Version". #buildinpublic #opensource
7
7
191
Did you know Google researchers developed a framework for AI-assisted peer review? There’s an open-source community implementation you can run inside Claude Code, Cursor, Codex CLI, and other coding agents. 🚨 Meet Open ScholarPeer. Based on Google’s ScholarPeer research, it breaks the review process into a structured workflow: > extract the paper’s claims, methods, and evidence > search for relevant literature > map how the field has evolved > hunt for missing baselines and datasets > run questions against the paper’s claims > produce a draft following the target venue’s review guidelines The useful detail: intermediate outputs are saved as Markdown files, so you can inspect what each step produced before accepting the final review. The project packages the workflow as skills, commands, and MCP tools that install into your paper’s project directory. You still check the evidence, edit the draft, and own the final judgment. An interesting way to bring a research review workflow into the AI tools developers already use. Free and open-source. Repo link in 🧵↓
32
20
1
122
10,892
Half the bugs I fix in the morning, I couldn't see at midnight. Sleep is still the best debugger I have.
10
14
225
Designer brain: does this feel calm? Dev brain: does this pass the tests? I need a yes from both before I ship.
1
5
95
Good morning, builders ☕ One day left. The launch video took 7 versions. Version 1 was a feature list nobody would finish. Tomorrow: peekover, my open source markdown reader. Free on Chrome and Firefox. Today: last checks, some small improvements. And not touching what works. #buildinpublic
3
1
10
260
Every "quick fix" becomes the most permanent code in the repo. Same with the furniture you were going to replace next month.
4
4
201
Do you remember Dreamweaver? #webdev #webdesign
2
3
53
Took me way too long to learn that shipping is its own skill. I was only ever practicing the building part.
6
5
96
Designers who learned to code: what was the first thing that almost made you quit?
2
76