Arjun Malhotra
@BadCapitalVC
Investing in companies that don't make money @GoodCapitalVC
1.4K Following 11.3K Followers
Reddit works as a GTM channel precisely because it resists being used as one. User trust only holds when the marketing on the platform is invisible.
intelligence should not be treated as a tool subscription you buy but as capital you allocate. a few ways to optimise for it:
1/ budget it by the job, not the tool. ask what a specific task is worth if a model does it well, and then what it costs to run. you'll then be able to fund the ones that pay back.
2/ match the model to what's at stake. if you don't put your best people on low-stakes work, think about intelligence the same way: a cheap model for the high-volume tasks and an expensive one only where a mistake is costly. most teams overpay by running a premium model across everything.
3/ don't over-commit all at once, as intelligence typically gets cheaper every few months. a task that isn't worth automating today often is 2 quarters later. underwrite in short cycles rather than locking in one big flat contract.
@stripe also put it well in their openrouter letter: you need to reason for the cost and return of every unit of intelligence the way you would for capital.
Show more
i think theres a larger picture to this
we might be underestimating the power behind and sitting down and writing/typing
wispr flow solves for the mundane and boring but not the fun and exciting, it seems almost too “dystopian post modern” to my liking as i prefer people being people rather than automated words, even typing mistakes tell a lot about someone lol
that being said, tech guys will know this is replicable but its <2% of the world, others depend on voice a lot so there is a market
Show more
yes @WisprFlow is a vitamin, not a painkiller. but it's defensible for plenty of reasons:
it's a product with a very high-frequency use case, even up to 40x a day. there's no learning curve to adopting it, and it integrates seamlessly with your current workflows. and distribution has been incredible, driven heavily by word of mouth.
subconscious adoption is a very underrated moat.
Show more
I don’t know how Wispr Flow is a business in 2027.
Literally weeks away from not needing to exist.
I have never seen a clearer example of Feature Not a Product.
It's interesting how when we talk about AI agents, we talk about the human being in the loop. But the human is actually the gap in the loop.
An ideal agentic loop runs closed; to involve a person, it has to halt, freeze state, render it legible, pass it over, and sit there waiting before it starts again on a changed input.
Despite this latency, I do think human-in-the-loop is a necessary evil for AI agents. Even trained on our own context and history, our internal agents land wrong roughly 2/10 times. And that's fine when the agent is drafting something, but not when it's set to do a task.
So ideally, every human-involved checkpoint in the agentic process should buy you accuracy if it's costing you time.
Show more
We're trying to run as much of the firm on AI as we can internally. So I thought I'd share where my own setup has gotten to:
My day was spread across email, calendar, WhatsApp, Granola, to-dos, trip plans and most of the friction was just me moving between all of them, everyday. So we built a dashboard that pulls from all of those sources into a single place.
It works like a card feed, and I swipe right if an action is done, left if it's deferred, and I can add notes if something needs to be picked up by the team. It refreshes 3 times a day: morning, midday, and end of day, as real time is expensive and I don't really need it.
The triage rules are simple: LP comms are higher priority, cold inbound goes to the team, warm intros come to me, newsletters get routed to my readwise.
Interestingly, the model turned out to be the least important piece when we were building this. There are 4 custom layers underneath: memory, context, skills and tools. The model sits on top and is basically swappable. Everything that took us months to get right lives in those layers, so switching from one model to another is close to a one-line change.
There's plenty still to figure out. Right now it's running at something like 2 in 10 false positives, so there's always a human in the loop and nothing goes out until I've approved the draft myself.
Even with that, it's kind of insane how much time it gives back. I'm not switching between 6 apps a day anymore, and that alone was worth the build.
Here's a dummy reference:
Show more
Been seeing a lot of speculation about MDR coming to UPI soon. Quick context, MDR is the small cut taken on a digital payment, split between the app, the network and the bank. Cards typically run 1-2% but UPI has been deliberately set to 0 since 2020 to get small merchants onto it.
PhonePe would benefit most if it's implemented. It runs almost half of all UPI value, 82% of its revenue is payments, right now and it can barely charge for any of it. A company this dominant is still losing money (with around $160m in the first half of the year) while Paytm made a small profit on similar revenue.
The timing makes it very interesting to me. They're filing to go public soon near $15 bn. Add even a modest MDR fee and they could be profitable as the expensive part of building the distribution is already done.
Show more
How are better-for-you brands & glp-1 drugs both having their moment?
Indulgence has been rising as a value globally, but it's still the lowest-ranked value we hold. This gives both categories a market, so:
a) better-for-you repackages indulgence as something virtuous and b) glp promises to handle the fallout
Show more
Not just pitch decks - even in writing, presentation, arguments, debates, simplicity is really the true measure of understanding.
complexity in a pitch deck is often inversely proportional to conviction
If you’re trying to understand the dynamic of real world agent adoption this post is a great place to start.
Everyone got so hooked on talking to chatbots that there’s limited recognition still that working with an agent is much more like managing someone in a process vs. just asking an ai some questions and getting a response back.
“prompting an agent is closer to writing a spec than asking a question. you have to scope the task extensively and define what "done" looks like.”
Ultimately, the real upside of agents is when you start to change the underlying workflow itself instead of just treating it as another system you ask questions of.
This means getting the agents the right data to work with, crossing organizational boundaries, and evolving the human in the loop steps for when people actually review the work. All of this has to change about today’s processes for the big upside to occur.
The end result is that it’s most likely that the vast majority of token usage in an enterprise will be agents that are “deployed” to go execute tasks inside of workflows.
Show more
"lastly, there isn't a job-to-be-done the public actually feels yet. autonomous agents will always be a solution looking for a problem."
spoke my mind - i was about to say the exact same thing
some obvious & non-obvious reasons i think AI agents may not have really been widely adopted yet, even though the tech is ready:
1/ it's not prompt in, answer out. an agent is a process you set up and steer while it runs, and the chatbot muscle memory most people have doesn't transfer.
2/ prompting an agent is closer to writing a spec than asking a question. you have to scope the task extensively and define what "done" looks like.
3/ as @paraschopra puts it, this needs a lot of delegation, which is a hard soft skill to build. it's very close to managing an employee and most people have never done this.
4/ a lot of the actual power still lives inside codex or claude code which is terminal-shaped and a little technical. you have to be comfortable doing the messy setup, so it self-selects for a narrow crowd.
5/ one agent is also just a tool. the unlock is running several at once and getting them to talk to each other like a team, and that handoff between agents is still mostly diy.
6/ same problem across people. your agent's context has to reach your colleagues or everyone ends up working in silos, and right now that handoff is way too manual.
7/ trust is a ratchet. a chatbot that's wrong wastes 10 seconds, but an agent that's wrong sends the email or edits the file. the downside is asymmetric, so most people keep it on a short leash.
8/ lastly, there isn't a job-to-be-done the public actually feels yet. autonomous agents will always be a solution looking for a problem.
Show more
Hot take… isn’t it kinda crazy that nobody is really using AI Agents? I don’t mean software engineers or AI early adopters. I mean “college friends talking about it in group chat,” the feeling you got when everyone started using Instagram or TikTok.
These frontier AI models are *insane* (as are the harnesses & tool calls & the like). And every large tech co has an AI agents platform, not to mention all the YC startups doing vertical agents. Yet all of your friends and family outside of tech — who spend all day staring at their iPhones and get paid to work in browser tabs — don’t really care or find themselves using any AI agents yet.
Yes ChatGPT, Claude, etc. are extremely popular… but if you look at the engagement data the vast majority of people are still using these aI chat tools like a glorified Google + Grammarly. That’s why the AGI labs are all pushing desktop apps for Codex, Cowork, etc. so hard to non-technical ppl. And yes exceptions for lawyers and customer service but even those have some asterisks and exceptions to rule.
Look I’m not saying the ChatGPT moment for AI Agents is not coming… it most definitely is! Remember we pivoted from Arc to Dia precisely because we believe computing is going to be radically reimagined around these AI primitives. No doubt. But that’s my point: it’s just so surprising it hasn’t happened yet because all of the tech you’d need is there.
Again if you stop for a second and think about it… for all the press and money and hype and models and crazy ARR numbers… this “AI Agent” moment does not *feel* like the other breakthrough tech moments we’ve lived through (e.g. think the shift to Stories via Snapchat & Instagram, or shift to on-demand via Uber/Airbnb/Doordash).
Which is a long way of saying: if you can figure out the answer to “why” most people don’t care about AI agents yet (and have no enduring interest in using them) — especially since the models and harnesses are here and ready — the answer to that question will allow you to capture a lot of marketshare and make a lot of money in 2027.
Theoretically, the tech is ready for AI Agents to totally transform how we work and live our lives… but alas the general public dgaf… that’s the generational puzzle to solve for the next 12 months for anyone not working on the models themselves.
Show more
interestingly, airtable actually saw the AI wave coming and tried hard to stay ahead. since 2021 they've been relaunching around ai workflows, prompt-to-app and standalone agents. but they lost the battle to horizontal & general-purpose agents on execution speed.
Show more
We've officially agreed to acquire Airtable for $1.285B! 😍
contrary to popular opinion, writing is still the proxy for us to determine how a founder thinks, even though everyone's using the same claude models.
there's a lot more slop, sure. but it's also gotten much easier to spot the slop.
Show more
Verifying written code is still the bottleneck, even though the cost of writing it is negligible now. As the OpenAI team puts it: the people who used to be bricklayers, handwriting every piece of code, now spend their days as construction managers.
This is a harder problem than it sounds, because a huge influx of written code brings with it a fraction of bugs and security holes baked in. At human pace, a 1% bad rate is noise you can catch. At 100x or 1000x the volume, that same 1% becomes a flood of problems moving faster than anyone can check.
How the OpenAI team deals with this is interesting: instead of one checkpoint at the end, they filter at many points along the way: one agent writes the code, a separate specialist agent reviews it for its own domain, another watches how it behaves once it's live, and so on.
Now that you can ship code at the speed of thought, I think the much harder part is fully trusting what you build.
Show more
Thesis might be the most misused word in venture. We use it to think of ourselves as forward-looking.
But most of what we call a thesis is just hindsight - usually we're just reverse-engineering the pattern after watching it play out across a dozen companies.
Show more
Swiggy Instamart now trails Blinkit & Zepto in AOV, dark store density and throughput. It's still deep in the red at around -15% EBITDA, while Blinkit has already turned positive. Even Noice, its premium private label, hasn't closed the gap on its own despite the products being great.
Their new survival move is to stop renting the marketplace and start owning the inventory, ie buy and hold stock, source directly from manufacturers, and control the shelf.
We've been thinking a lot about supply-first businesses, and owning the assets is one of the clearest ways to be supply-first. When you hold the stock, you can be strict on quality, availability, and fill rates- none of which you can vouch for when a third-party seller sits in the middle.
And it also changes the margin architecture: you capture a slice of the retail value chain. It's a harder model to run, but if they're successful, it's also a much harder model to copy.
Show more
was talking to a founder recently and realised how all these businesses AI is supposed to kill (HR consultancies, telecalling, services firms) won't go anywhere fast.
they've got relationships & clients that buy them years. but that same trust is what lets them coast. no reason to fix the backend, retrain anyone, rethink how the work gets done. so the ground shifts under them slowly and they die a slow death anyway.
i think we'll see a lot of operators not competing with them but buying them outright, rebuilding the ops underneath with AI, and reselling a leaner business. kind of like buying a distressed asset, except the distress here is technological.
Show more
If you're selling rooftop solar to Indian homeowners, trust is the only thing that gets someone to put hardware on their roof for 25 years.
Everything else - financing, install quality, service - is downstream of whether the household believes you'll still be around in a decade.
We've been backing SolarSquare since their early days. 50,000 homes across 29 cities in, and they haven't slowed down on building that trust once - which is what makes MS Dhoni signing on as an investor & brand ambassador such a great fit.
Welcome to the captable, Captain!
Show more
