I build AI-native products, apps & automations. Writing about tech, AI and things that industry refuses to admit | Feature your brand: →brandmysetup.lol
0-100
Based in India
- Tweets 9,894
- Following 839
- Followers 20,921
- Likes 27,921
Got to know Google doesn’t give a fuck about vanity AI leaderboards.
They are moving towards what actually matters: Profitable AI Infrastructure
2
681
Xiaomi just dropped MiMo-V2.6-Pro and casually jumped past GPT-6 Luna, DeepSeek V4.1 and GLM-5.3.
What the hell is happening?💀
5
693
I finally started posting on Instagram.
Meet SpiderCave 🕷️
Going to be sharing more setup POVs, aesthetics, and behind-the-scenes stuff there.
16
2
73
1,745
C’mon Claude do some mathematics.
Terence is going to lose his mind again.
Mathematician Terence Tao:
"we have to slow down AI. the pace is insane, and there's no reason to be this fast — no reason at all"
It's amazing how willing we are to change everything without any idea what happens afterward
These are extremely nonlinear dynamics
6
924
Here’s how these guys managed to hack openAI using claude Opus 5.
Harsh, Mohan & Rahul from hacktron AI were investigating OpenAI’s security as a part of vulnerability research.
1. They started with an image upload
OpenAI’s community forum processes the HEIF/HEIC images. The researchers discovered a vulnerability in the image-processing chain involving libheif. A specially crafted image could trigger a memory-corruption bug and potentially execute code on the server.
2. They brought in Claude
So they used claude to help turn the vulnerability into a working exploit.
Claude Opus 4.8 struggled with it.
So they went for Opus 5and gave it the same problem and reportedly got a working exploit within hours.
3. One vulnerability wasn’t enough
Finally after getting access to the forum environment, they discovered another weakness in OpenAI’s authentication infrastructure. That allowed them to obtain auth tokens and take over ChatGPT/Codex accounts belonging to OpenAI employees.
4. Then came the crazy part
One compromised employee’s Codex account was connected to OpenAI’s private GitHub organization. That gave the researchers a path into OpenAI’s internal software environment.
5. They proved it without stealing the code
Instead of downloading the private source code, they used the compromised Codex account to create a harmless pull request in OpenAI’s internal repository.
That was their proof of access. Then they stopped and reported the vulnerabilities. The entire chain took less than 72 hours.
And finally OpenAI fixed the reported issue and awarded the researchers a $6,500 bounty.
Here are the LinkedIns of the 3 Indian guys who hacked OpenAI with Opus 5 in 2 days for <$3000:
Rahul Maini, Mohan Pedhapati and Harsh Jaiswal
No brand name colleges or companies. Just raw curiosity and skill.
You can just do things.
4
14
1,506
His career arc goes absolutely crazy.
> Started in investment banking
> Moved into product
> Built Cube26
> Got acquired by paytm
> Launched pocket FM
> What a fucking trajectory
Introducing Sherpa: the most advanced fiction writing AI
We accelerated from $250M in ARR to $500M because Sherpa helped increase content production by 1200% in 1 year
Sherpa was trained on 5.5B hours of playtime with minute by minute dynamic retention data.
550K+ creators have produced 2.6M hours of content annualised using it
Pocket FM is like Netflix for audio-only dramas, with our own pool of one-person studios.
10% of eligible writers on Pocket FM make >$200K
One blockbuster produced >$100M in revenue
3 writers have become millionaires in <2 yrs
We built Sherpa to enable anyone to make >$1M by writing world-class fiction stories:
1. The Idea: Drop a 1-2 sentence concept. Sherpa interrogates it like a veteran editor on tension, stakes, and psychology
2. World & Characters: It builds out the complete lore, tone, and character psychologies
3. Sub-Plot planning: Breaks the premise into arcs, arcs into episodes, and episodes into scenes
4. Scene-by-Scene Generation: Outlines and drafts entire episodes, with you able to steer, rewrite, or override anytime
5. Editorial Review: Stress-tests every draft for pacing, engagement drop-offs, prose, and coherence before it locks
6. One-Tap Production: Pick a voice, convert to audio drama, and publish directly to Pocket FM’s millions of listeners
7. Global Scale & Monetization: Revenue-share on performance, with automatic localization so you earn across international markets
Test Sherpa for free here: pocketfm.com/sherpa
_____________________________________________
Generic LLMs fail at serialized fiction because they lack a long-horizon narrative reward function.
Sherpa solves this through three core technical leaps:
1. Narrative World Model (State Tracking & Retrieval): Context windows degrade over long runs. Sherpa constructs an evolving semantic knowledge graph tracking character states, secrets, and plot dependencies. High-speed retrieval surfaces exact context on demand, maintaining zero continuity decay across hundreds of episodes
2. Hierarchical Story Planner: When writing a 500-episode story like Naruto, you need to plan 100s of sub plots. Rather than generating linearly, Sherpa decomposes narrative across discrete levels: season -> arc -> sequence -> episode -> scene.
Rather than generating everything upfront, like a generic LLM, Sherpa uses progressive planning and dynamic replanning. As the story evolves, it identifies what changed, traces the downstream impact, and replans only the affected parts.
3. Prose Engine (Trained on series' retention data): LLMs write robotically, but serial fiction needs emotion, tension, pacing, and dialogue that sounds like real people.
Sherpa's Prose Engine was designed specifically for storytelling. It was built on 1B+ tokens of Pocket's own stories, trained by learning from what listeners engage with, where they drop off, and what keeps them hooked.
Feedback is taken from specialized evaluator models that measure every scene against a 40-item checklist. (Evaluator models were benchmarked against human reviewers and matched them 80–90% of the time.)
_______________________________________________
Owning distribution and creation puts us in a very unique spot.
More shows -> More data -> Sherpa becomes better -> more creator success -> more creators -> more shows
Pocket FM has already seen one $100M IP. I believe Sherpa will soon lead to dozens of single-person studios creating billion-dollar shows.
Most people are scared of AI but I think it'll unlock more human creativity, help creators earn more, and bring the next great IPs to life. This will create millions of jobs and new income streams.
1
11
775
I built a marketplace for brands to rent space on my desk setup.
This is probably the weirdest thing I’ve built this month.
Well, saw a bunch of new visitors appeared this morning.
I'm selling ad space on my Setup.
→ 12 ad spots across the setup
→ 4 biddable & 8 non-biddable spots
→ 7-day live auction periods
Your brand posters gets featured on my setup and appears in my videos and posts across social media.
→ brandmysetup.lol
1
7
931
I’m selling ad space on my desk setup.
- 12 ad spots
- 4 biddable & 8 non-biddable
- 7 days live auction
Checkout: brandmysetup.lol
1
3
449
The 3D setup was cool.
But here’s what it actually looks like IRL
You can literally put your brand on it.
-> brandmysetup.lol
I'm selling ad space on my Setup.
→ 12 ad spots across the setup
→ 4 biddable & 8 non-biddable spots
→ 7-day live auction periods
Your brand posters gets featured on my setup and appears in my videos and posts across social media.
→ brandmysetup.lol
9
59
3,091
I'm selling ad space on my Setup.
→ 12 ad spots across the setup
→ 4 biddable & 8 non-biddable spots
→ 7-day live auction periods
Your brand posters gets featured on my setup and appears in my videos and posts across social media.
→ brandmysetup.lol
I'm selling ad space on my MacBook
→ 10 sticker spots on the lid
→ live auction
→ 14 days
except I don't own it yet, the auction buys it
→ your logo travels with me and gets seen both online and IRL
BrandMyMac.com
26
4
84
12,989
AI Startup founders calculating gross margin after discovering their agent has been calling GPT-6 for 2 hours.
1
1
3
853
Don’t waste your entire context window on video inputs for agents.
Someone open-sourced a surprisingly elegant way to give LLMs video input.
It samples the important frames, detects scene changes, keeps track of token costs, and builds the smallest useful context.
2
2
799
Mark Lou’s story is far more inspiring than his $143K/month.
But because before that milestone, he failed, cried, started over, got rejected, moved countries, and even broke down.
Let’s deep dive into his journey.
Early 20s:
Mark had no idea what he wanted to do with his life just like us. He was addicted to video games, had lost his confidence, and wanted to prove there was another way to build a career.
So he started building tiny internet businesses. He learned everything from YouTube coding, marketing, even English. He was a french guy & he could barely speak english at that time.
First attempt:
He spent an entire year building a mobile app. The problem is he barely understood how the code worked.
He was copying pieces of code from libraries, adding features, and hoping everything would somehow come together.
But It didn’t. The app became a buggy disaster. Then an entrepreneur asked him one simple question:
“How are you going to make money?”
And he has no idea what to answer.
2017–2018:
Soon after Mark moved to South Korea with his girlfriend and lived in a tiny underground apartment. It was cheap, humid, and full of insects.
He took random gigs to survive while trying to build software for businesses. Eventually, he got his first customer.
Then another. Then another.
His process was painfully simple:
1. Find a problem
2. build something
3. contact people
4. improve it
5. repeat.
2018–2019:
The business started growing. Mark and his girlfriend moved to Bali again.
He surfed in the mornings, worked on marketing, and coded in the afternoons. By the end of 2019, the business was doing around $4,000/month.
For the first time, it felt like his crazy experiment might actually work.
Then suddenly 2020 happened.
COVID changed everything. They returned to France and got stuck there. His previous software started losing customers. The money disappeared.
And Mark eventually found himself living with his parents again.
Late 2020:
At 27, he was back where he started.
Now he was sitting in the same house with failed projects, little savings, and no idea what to do next.
Finally he hit his lowest point. So he made a decision he never expected to make: got a job.
Applied for software engineering roles while teaching himself APIs, Node.js, React and everything else he didn’t know.
he was rejected because he wasn’t really a software engineer yet. Eventually, one company gave him a chance.
But the job wasn’t the end of the story. Basically it was the reset.
2021:
Mark decided he would approach building differently.
He created 5 rules for himself:
1. Sleep properly.
2. Build in public.
3. Ship the smallest version possible.
4. Launch everything.
5. Focus on process instead of outcomes.
No more spending a year secretly building something before showing it to anyone.
2021–2022:
He started shipping tiny websites.
- Mood to Movie.
- Habits Garden.
And many others.
Some failed. Some made a little money.
But one of them unexpectedly went viral on Hacker News.
Habits Garden eventually reached around $500/month in recurring revenue.
Nothing was life-changing yet. But something more important was happening: He was getting better at building.
2022:
By the end of the year, Mark had built around 10 websites. Most weren’t huge successes.
But every one taught him something.
- Coding.
- Marketing.
- Distribution.
- Pricing.
- Product.
And he was sharing the entire journey publicly. Then his best friend passed away.
It forced him to rethink what he was chasing. He realized the value wasn’t always in the outcome.
Sometimes the value was simply in the process. So he kept shipping.
2023
Mark changed his strategy entirely.
Instead of building whatever sounded interesting, he started looking for real problems. Painkillers, not vitamins.
He built Indie Page.
Then he built MakeLing during the AI boom. but he quickly realized he didn’t care much about the market.
More below 👇
12
9
1,463
