This is why we are building HyperTest
PROBLEM:
AI is the future of software development
@Replit@cursor_ai@windsurf_ai@boltdotnew....
▶️ Copilots producing code, at 2x-3x speed
▶️ But with 40% more design & security flaws
▶️ One broken backend call breaks all connected calls instantly — AI errors spread like 🔥
▶️ QA can’t keep up — AI will break app 5x-6x faster than human testing can scale 💔🔨
A lot of analysts predicting AI agents automating human tasks in 2026 should temper their expectations seeing what Salesforce saw
1. AI 'drift': When users ask irrelevant questions, AI agents lose focus
2. Reliability: Despite providing clear instructions to send satisfaction surveys agents sometimes forgot to send them for unexplained reasons
Most models suffer from lost-in-the-middle or attention dilution issues i.e. when asked to act on a set of instructions sometimes 8-10 steps long they inconveniently forget a few.
https://t.co/vj9x9qqRye
ESOPs look worthless because employees (including vested founders) stand in the line behind investors with liq. pref. I guess everyone knew a fat cheque at the end was not a certainty.
5 years and now my ESOPs are worth practically nothing. What an outcome!
Startup ESOPs means nothing. Never sacrifice your base pay for some paper money. Lesson learnt the hard way.
Here is a few things I'd love to understand
1. I login in my app using an OTP that I get on my mobile. So when their AI loads my app and they would press the login button how would they punch the OTP to move to the next page?
2. I add products to my cart before I checkout. if their AI gets to the product page and try adding a product which is out of stock how do they get to the cart page?
3. I apply discount to my cart and checkout. If their AI applies a discount code and see it is no longer valid or no longer valid for the product or for the day how would they verify if my discount feature is working right?
There are just so many other scenarios that where the AI not in control of the db state would stutter in a flow.
How does 'db branching' handle creating the right db states?
Introducing Anything Max: Vibe Coding that's leaps above Lovable and Bolt
We've raised money at a $100M valuation and built what we believe is the future of vibe coding.
We asked 100 vibe coders to build their apps side by side on Lovable, Bolt, and Anything Max and they rated Anything Max the winner across all 3 categories - accuracy, design, and 'overall'.
Here's why:
• Full-stack control: Max can test backend hooks, branch database states, and debug issues, because Anything owns the full infrastructure.
• Max can load up your app in its own browser and click on all buttons like a human tester to find all edge case bugs, then trace the bug across the stack - could be a frontend, backend, or a database issue (only we can do this, read #1) and autonomously fix it with 97% accuracy.
Lovable and Bolt build prototypes, but Max users are building production-ready apps and already charging money for them.
Blake built a gut biome app to $10K run rate
Anthony built a referral tool to $20k in revenue
Yuri built a suite of apps doing $40K
Build your app with Max: https://t.co/haQw8s04mB
--------------------------------------------
We're hosting a $100K Hackathon to help people grow their app to $10K MRR.
- We'll teach you everything we know about growing to 1M users.
- You'll have 30 days to build a real product in public and get paying customers for it.
If you do it well, you can start the New Year with a functioning business.
Retweet and comment “LFG��, and we’ll send you a $100 discount code and the link to participate
Bootstrapping forces you have to ship fast, launch small, and get real feedback early. You simply cannot spend 6 months or a year building in the dark and hoping for a perfect (expensive) big launch.
When we were finding the first 1000 users for @TallyForms , we didn’t optimize for MRR growth. We focused on hyper-targeted (cold) outreach to get feedback from the right early adopters. The people who loved what we were building even in its earliest form. For us those were founders, product teams and @IndieHackers.
Those early believers helped us refine, iterate, and build something that resonates. We asked feedback to thousands of people manually for 8 months and invited them to our Slack. Those conversations shaped the product far more than any big launch could have. And optimizing for them turned them into our biggest ambassadors.
Most PM, engineer turn founders neglect list building, cold emailing, cold calling for the longest of times.
They think they should get the product right and then get someone to build those lists, write those copies and things will fall in place.
but building those lists, and writing those copies is one half of getting the product right.
Every time I'm at a YC event, I get asked how to do GTM.
My answer is always the same - here are the bullet points:
1. build your lead lists, use tools like Crustdata or Apollo
2. don't use AI to create the outbound email - make it as concise as possible
3. take a look at deliverability metrics - iterate email every day - make it more and more punchy
4. separately, use LI Sales Nav to write 2 sentence outreach to people - preferably with warm connection (common university, etc)
5. Post on LinkedIn everyday
6. invest in SEO
spend 80% of your time on the above and you will breakthrough and get demos
Every founder who followed this advice with enough discipline saw results.
Just shipped HyperTest to Harness.
Jyoti built the company I admired. Now harness is using HyperTest.
SaaS founders before me and many after look upto him to believe what is possible - young Indians moving to valley. Building without fear. Seeing success.
Getting it right with Appdynamics. Then again with Harness
That's the moment you realise what would be the right way to reach out to people you look up to. when you want advice. when you want to pitch them. Maybe just to build something they'd actually want to use.
Everyone's out there waiting for the perfect time to connect with their heroes. Turns out the perfect time is when you've built something worth their attention.
Now back to building 🧑💻
iii) Citation-ready over engagement optimisation
Are your key findings easily extractable? (AI needs to quote them)
Are your sources clearly marked? (AI needs to trace where you got information)
Is the claim structure logical? (AI needs to understand what you're actually arguing)
iv) Creating "citability scores" for your pages
"This page has 8/10 citability (clear methodology, original data, good sources) vs. this page has 2/10 (it's commentary on other people's work).
AI agents are changing the way outcomes are delivered. Web search that controls how, where and when our users discover and engage will shift dramatically.
With all large model companies now launching their own browsers it is clear in future agents not humans would be searching web
This means marketers who harness traffic, ranking and visits would adjust for how agents browse web.
Read on >>
6) i) consolidation over proliferation
You used to think: "More content = more keyword targets." Now you realize: "More similar content = AI confusion + no citations." You'd actually delete or merge pages rather than create new ones.
ii) Authority signalling over keyword matching
You shift from: "Does this page have the keyword?" to "Does this page clearly demonstrate authority?
>>
What exactly do you get by cold emailing folks "2-levels above your league" Founders are constantly cold mailing potential customers and getting rejected to have our hustle in check.
Most times users don't know of the better world possible with your software and the constant grind is to somehow impress that to them in less than 60 words.
That is all that is needed. if the 2 levels above your league people who are not exactly your user why care reaching out
If you aren't cold calling/emailing 1-2 people per month who are 2-3 levels "beyond your league", and not getting rejected 50% of the time, you aren't hustling hard enough. START. If you are getting rejected 100% of the time, you haven't figured out how to cold open yet. STOP.
It helps to have someone head a venture fund or accelerator who meets young founders day in day out and pushes this to think big, never feel limited by boundaries on market and tech that others set for you.
but it is completely else to be delusional to the point that you misinterpret the impact of new tech that does not seem like pushing real boundaries but imagining completely absurd possibilities that violates the laws of physics.
case in point: Thinking of flying taxis to take you to Mars before you have seen take you from point A to B on earth
vibe coding is good at best at producing prototypes but fully functional apps that can take users, has multiple moving parts and a character of its own can not exactly be vibe coded.
All the garry needs to do is google search apps built with https://t.co/9bdYrcNxYm and lovable and he'd know.
If our business would be the first to be competed away by vibe coded apps, why are we seeing such rapid customer growth (exceeding 50%) right now? And why don't we see vibe coded email or spreadsheet or accounting app or messaging apps yet?
My own personal R&D project is to enable huge gains in programmer productivity by _combining_ compiler technology with AI. Our goal is to enable a quantum leap in programmer productivity while being able to provide security, privacy and compliance guarantees. Without those guarantees, vibe coding just piles up tech debt faster and faster until the whole thing collapses.
Of course, for people like Garry Tan, tech debt is to be pawned off on unsuspecting acquirers.
Let me make a bet with Garry Tan: we will outshine and outlast his vibe coding companies!
All apps made through bolt, emergent are static sites with some graphics and supabase / firebase. I 'd like to see folks using these tools to just connect a working database and I'd be mighty impressed!
Zoho’s business would be first to be competed away by people building their own custom software built by people using @Replit@emergentlabs and @Taskade
Why pay $30/seat/month for over bundled SaaS when soon even nontech ops ppl can vibe-code a custom solution in a weekend?
Give me one example of a successful agentic execution of workflow i.e. 4-5 steps to be done in sequence through a browser getting done with the same (not even better) outcome than a human
MCP browsers are no innovation.
Browser automation is quietly becoming the new execution layer for enterprise ops.
Most finance/HR/ops/compliance work now lives in SaaS + external portals, not APIs. LLMs finally make browser agents reliable enough to handle that long-tail. The winners will be vertical, with deep tool libraries, 20-40 pre-built workflows and auditability.
Huge, unsexy TAM (which is good). Feels like RPA 2.0.