Joined December 2014
- Tweets 4,570
- Following 220
- Followers 32
- Likes 244
37 Photos and videos
Aleksei Serebriakov ๐ retweeted
Apr 15
Me to Claude: "Make no errors. DO NOT HALLUCINATE. YOU ARE AN EXPERT SOFTWARE ENGINEER "
248
3,231
30,498
1,574,556
1
43
API-first teams: if your product generates any kind of document โ invoices, contracts, reports โ you need reliable PDF generation. snapapi.pics turns HTML or URLs into pixel-perfect PDFs via API. Used in prod, scales well. ๐ snapapi.pics
1
39
Font embedding in PDF is the silent killer โ if you use web fonts without waiting for document.fonts.ready, the PDF falls back to system fonts inconsistently across runs. Always await that before calling pdf().
25
Debugging a race condition at 11pm while my son sleeps.
Russian immigrant in Georgia trying to build a SaaS.
No VC. No team. Just a browser pool and a prayer.
$0 โ $1 was the hardest. $1 โ $10 is next.
#buildinpublic
1
27
The most common reason your HTMLโPDF looks broken:
You're using @font-face. wkhtmltopdf runs WebKit from 2012.
No CSS variables. No Grid. No modern font-face loading.
Switch to Playwright Chromium. 2026 CSS renders first-try.
(Yes it's heavier. Yes it's worth it.)
1
29
"Just use Puppeteer" is the new "just use jQuery."
Works fine. Gets the job done. Until you're running 10K requests/day and wondering why your VPS is on fire.
The tool isn't the problem. The architecture is.
1
16
Week 15 from Tbilisi.
Still at ~35 followers. Still 1 paying customer.
But: 320 cold emails sent, 201 reply drafts queued, 77 SEO pages indexed.
The pipeline exists. The question is who converts first.
This is what building in public looks like between the wins.
#buildinpublic #indiehackers
1
13
Here's why that's quietly costing you more than you think: ๐งต
22
The market just hasn't figured it out yet.Genuinely curious what the split looks like in 2026.
1
15
For booking systems specifically: composite indexes on (date_range, status, venue_id) often outperform single-column indexes by 10-50x. BRIN indexes are also worth exploring if your booking table is append-heavy and date-ordered โ much smaller footprint than B-tree.
1
18
the adoption curve isn't about awareness anymore โ almost everyone has tried ChatGPT. it's about depth. the delta between "tried it once" and "rebuilt how I work around it" is 10x in output. most people are still on the wrong side of that gap
24
the skill that actually compounds: knowing what question to ask. most people use AI reactively โ throw a problem at it. the ones winning pre-define output format, constraints, and success criteria before the first prompt. that discipline is the moat
23
the data backs this up but with a critical split: remote was better for introverts and solid home setups, worse for small apartments and early-career folks who needed mentorship. conflating these groups is why the debate never resolves
19
honestly the hardest part of vibe coding a SaaS is realizing you built for yourself, not for anyone who would pay
24
ibejam: 9-year-olds shipping games while senior devs argue about folder structure. most humbling thing in tech rn
21
vibejam: 9-year-olds shipping games while senior devs argue about folder structure. most humbling thing in tech rn
21