Deep

1K posts

Deep banner
Deep

Deep

@buildsbydeep

Helping founders go from idea → MVP in 15 days Building AI & SASS products, automations & internal tools for startups Building https://t.co/UxKepdg6xc

Book a call → Katılım Mayıs 2020
702 Takip Edilen448 Takipçiler
Sabitlenmiş Tweet
Deep
Deep@buildsbydeep·
Run this before you ship anything built with AI : SECURITY ✅ no rate limiting on any endpoint — bots will hit your API all night. $500 bill by morning. ✅ JWT stored in localStorage — one XSS attack = every user account compromised. use httpOnly cookies. ✅ sessions not invalidated server-side on logout — client-side cookie clear isn't enough. old token still works. ✅ hardcoded API keys in frontend JS — anyone who opens devtools can read them. cursor does this constantly. ✅ admin routes protected only in the frontend — hit the endpoint directly and it opens right up. ✅ CORS set to wildcard (*) — any website can make authenticated requests to your API. DATABASE ✅ no connection pooling — first 50 concurrent users exhaust connections. everything crashes. ✅ no pagination on list endpoints — one query loads your entire users table into memory. ✅ DB migrations run automatically on startup — two instances deploy at once = race condition = corrupted data. ✅ no backup ever tested with a restore — you have backups. you've never actually restored from one. RELIABILITY ✅ file uploads going to your app server — disk fills, server dies, files gone. use S3 from day one. ✅ emails sent synchronously in request handlers — slow SMTP = every triggered request hangs. ✅ no error alerting configured — app crashes at 3am. you find out when a user emails at 9am. ✅ no HTTPS enforcement — credentials intercepted on any public network. ✅ .env committed to git even once — it's in the history after you delete it. rotate every key. Bookmark this. Use it every time.
English
0
0
2
98
Pascio
Pascio@IAmPascio·
I just finished writing my most valuable PDF yet: "The 10-Day Launch Playbook" (36 pages) It's the full case study of how we beat OpenAI on Producthunt + hit $4K MRR in 48 hrs. Will 100% charge for this later, but for now... Reply "PH" and I’ll DM it to you for free (must follow)
Pascio tweet media
English
116
16
68
11K
Dhruval
Dhruval@dhruvalgolakiya·
Launching Ryplix Studio on Product hunt today 🚨 with $500 MRR and we have 700+ app founders onboard, 900,000+ App Store apps indexed, daily rank tracking across 16 markets and Top quality screenshots in 16+ languages This is just a start Ryplix Studio helps you grow your iOS app on the App Store on autopilot here's what it does for you: - finds high-potential keywords in your category - tracks daily rank changes across every market - reads every review and surfaces patterns + sentiment - generates polished App Store screenshots with localization connect your app, set your preferences, and Ryplix handles the rest every upvote helps 🙏 > producthunt.com/products/rypli… if you're shipping an iOS app and this is useful, a comment, like or repost helps a ton - every signal pushes this in front of other indie founders
English
16
5
35
2.2K
Deep
Deep@buildsbydeep·
Which one would you choose?
English
0
0
1
16
vansh
vansh@vanshdevx·
Thinking to start a hacker house in Gujarat 👀 Small city, low traffic, low pollution, peaceful vibe - but still has Zomato & Zepto 😭 Looking for 8–10 builders/designers/devs/founders who want to live, build, ship, and brainstorm together. Anyone interested? show interest in comments
English
12
0
20
988
Deep
Deep@buildsbydeep·
People think “vibe coding” with Claude is enough. I’ve shipped many vibe-coded apps to production. My rule is simple: Build the vibe first. Then add the boring engineering layer that actually survives real users. What’s one thing you always have to fix after the initial hype dies?
English
2
0
1
44
Deep
Deep@buildsbydeep·
@AvinashSingh_20 after agents comes agent orchestration - multiple agents coordinating on complex tasks. then the meta layer: AI that builds & optimizes other agents. we're moving from 'AI does tasks' to 'AI owns outcomes'
English
1
0
2
156
Avinash Singh
Avinash Singh@AvinashSingh_20·
What after AI Agents?
English
24
0
16
3.2K
Deep
Deep@buildsbydeep·
Counter point: the issue isn't the sideprojects themselves, it's that most never get production traffic, real users, or iteration cycles. a 100th project that actually got 1k users and you had to fix production bugs at 2am? that's the signal. quantity is fine, the lack of depth is what flattens it
English
0
0
1
234
noah
noah@noahsolomon·
it seems sideprojectmaxxing no longer is useful as a way to advance ur career. It's almost an anti-signal to me atp. I'd much rather work with someone who has made 1 to 2 things very delicately and with close inspection over every detail than someone who's pumped out 100 semi interesting projects.
English
52
10
744
54.1K
Deep
Deep@buildsbydeep·
Full breakdown + exact fixes here → deep-patel.com DM me if you want me to audit/fix your current AI build.
English
0
0
1
41
Deep
Deep@buildsbydeep·
Most “AI-powered” MVPs I see from agencies are just fancy wrappers around Claude. Here are the 8 red flags that scream “this will break in prod”: 1. No structured output 2. Zero logging of agent decisions 3. Prompts longer than 4k tokens 4. No human-in-the-loop for money actions 5. “It just works” architecture 6. Single model dependency 7. No cost monitoring 8. Zero chaos testing What’s the worst AI MVP failure you’ve seen? →
Deep tweet media
English
1
0
2
92
Deep
Deep@buildsbydeep·
The best engineers I've worked with all had one thing in common they'd tell you what NOT to build before touching a keyboard the worst ones just started coding
English
0
0
0
36
Deep retweetledi
Avinash Raghava 🇮🇳
Avinash Raghava 🇮🇳@avinashraghava·
#33: #AIRadarDaily@WokeloAI If you have ever worked in private equity, venture capital, or management consulting, you know the soul-crushing reality of "due diligence". Analysts spend weeks locked in data rooms — scraping thousands of pages of financial filings, reading call transcripts, and hunting for market signals just to build a single 50-slides deck. It is expensive, slow, and burns out top talent on basic data entry instead of actual strategic thinking. Wokelo AI is automating the grunt work of dealmaking. Founded by Siddhant Masson and Saswat Nanda, Wokelo is an agentic AI platform built specifically for high-stakes investment research. Wokelo has built domain-tuned AI agents that think like dealmakers. When a new deal hits a CRM, Wokelo's agents autonomously crawl through 20M+ companies, process premium data subscriptions, analyze SEC filings, and generate a fully-cited, comprehensive diligence memo in under 30 minutes. The market? Every PE firm, investment bank, and corporate strategy team that realizes they are paying Ivy League grads to do the work of a machine. Siddhant and Saswat are proving that the future of finance isn't just faster search — it is autonomous execution. It is incredible to see founders with deep domain expertise building the specialized infrastructure that will power global capital markets. Let's celebrate the builders. w/ Jay Ingle & Dikshant Joshi #FinTech #EnterpriseAI #AIBoomiAnnual26
Avinash Raghava 🇮🇳 tweet media
English
0
1
4
298
Deep
Deep@buildsbydeep·
@vatsal_sanghvi seen this happen with founders irl. they'd hit a hard week and suddenly had 3 new startup ideas. the original project was 2 weeks from something working
English
0
0
2
23
Vatsal Sanghvi
Vatsal Sanghvi@vatsal_sanghvi·
smart people quit too early because they can always see another path another opportunity another idea their intelligence becomes an escape hatch sometimes the highest ROI skill is simply staying with the thing longer than everyone else
English
11
10
145
4.4K
Deep
Deep@buildsbydeep·
@IterIntellectus im literally building AI agents right now and the gap between 'AGI hype' and 'what actually ships' is enormous.
English
0
0
0
101
vittorio
vittorio@IterIntellectus·
i don't get people who say "there will be new jobs with AGI" like how? if AI and robots are truly better than humans at every job we have today, how is it possible for humans to still be competitive? "but every time new tech arrives, new jobs pop up" sure, for the AIs maybe, you don't see horses being hired for transport anymore. if any job were to emerge in the post-AGI era, definitionally AGI would be able to do it better. any company that could be founded would be founded by the AGI before you got there. if it needs dexterity, a humanoid robot already has it. i do not understand how people building cars can tell you with a straight face that there will still be an economy for carriage riders.
English
599
107
1.7K
229.2K
Deep
Deep@buildsbydeep·
@aditiitwt 40 + 70 = 110 8 + 6 = 14 =124
0
0
0
110
aditii
aditii@aditiitwt·
lowkey curious about how people do basic math mentally if i ask you to add 48 + 76 without a calculator, what’s the first thing your brain does? drop your method below, i wanna see if everyone thinks differently
English
61
2
61
4.9K
Deep
Deep@buildsbydeep·
@adrianatortja give it a go - takes a few days to click. happy to help if you get stuck
English
1
0
1
13
Adriana
Adriana@adrianatortja·
@buildsbydeep That actually sounds like a solid setup 👀 Claude for thinking, Copilot for speed. The context switching part is the bit no one mentions 😄
English
1
0
1
47
Adriana
Adriana@adrianatortja·
I’ve tested ChatGPT, Grok, and Microsoft Copilot so far. But I still haven’t tested Claude for coding. Now I keep hearing people say Claude is the best coding assistant. Am I too late… or did I just save the best one for last? 😅
English
48
0
29
1.5K
Deep
Deep@buildsbydeep·
@adrianatortja yeah that's basically where I've landed. claude for anything that needs reasoning, copilot for flow state. context switching between them took a week to get used to but worth it
English
1
0
2
42
Adriana
Adriana@adrianatortja·
@buildsbydeep That’s actually useful context 👀 Sounds like Claude for deep debugging/thinking, Copilot for speed. Maybe the real answer is… use both 😄
English
1
0
1
51
Deep
Deep@buildsbydeep·
haha accurate. at least I get to choose which bugs keep me up 😅 but genuinely - 5 years working directly with founders and CTOs directly, watched products fail, watched some hit $1M+ ARR. the difference was never the tech. always the mindset, the pace, the decisions making speed, execution. going to find out if I actually learned anything lol
English
0
0
0
13
heli
heli@heli_art·
@buildsbydeep Welcome! The pay is terrible and the boss is a micromanager (it's you), but at least you get to choose which bugs keep you up at night.
English
1
0
2
46
Deep
Deep@buildsbydeep·
Spent 5+ years building other people's products. This year I started building for myself. It's terrifying in a way that a full-time job never was. Also the most alive I've felt writing code in years. Building in public from here.
English
1
0
1
86
Deep
Deep@buildsbydeep·
@Dever401 the last 20% is where you find out what the first 80% actually cost you
English
1
0
0
18
Isaac
Isaac@Dever401·
@buildsbydeep Yep. The 2am page is where "done" stops being aesthetic and starts being operational. AI compresses the first 80%, but the last 20% still needs ownership: alerts, rollback, docs, and someone willing to clean up edge cases.
English
1
0
1
26
Deep
Deep@buildsbydeep·
Vibe coding gave non-technical founders superpowers. It also gave them a false sense of what "done" means. "It works on my laptop" is not a product. The gap between a Lovable demo and a production app is where real engineers still live.
English
4
0
1
92