Amit | ReconPe

Finance teams want reconciliation. Not rule engineering. So we built a template marketplace — ready-made reconciliation playbooks for marketplaces, banks, and the GL workflows that show up in every close.

Selling on marketplaces is hard enough—reconciling your settlements shouldn't take days. I built ReconPe to automate the chaos. No more manual column mapping, no more "marketplace math" eating into your margins. Just drag, drop, and reconcile. Here is how ReconPe handles a month of data in minutes: Match Payouts: Instantly see which payments landed and which are missing. Detect Overcharges: Audit fees against rate cards to spot overbilling. GST Recovery: Automatically surface reclaimable input credits hidden in marketplace fees. Full Visibility: A clear waterfall report from Gross Sales to Bank Balance. Stop losing money to hidden settlement errors. Check it out at reconpe.com and take control of your e-commerce finances

@MKBHD Getting overwhelmed by the pace of production and features from all the big giants in this space. Seems like this is the new normal test, adapt and move on.

It is getting genuinely difficult to keep track of all of the names of AI products being unveiled. In the last hour, Google's unveiled Google Pics (which is not Google Photos), and updates to Google Flow, Nano Banana, Veo (all media generation), Google Antigravity, Gemini Spark, Gemini Omni, Gemini 3.5 Flash

Your LLM makes a confident mistake. You fix the prompt. Tests pass. PR merges. Three months later the same class of bug is back on a different field. Someone pieces it together from scratch. The fix went into git log. The lesson didn't. 👇 medium.com/ai-advances/tr…

I am a fan of @ylecun since this line, no complexity to the point Year 2024 - 4 year old child ~~ adjusting to new models 40 times more data than the biggest LLM

Yann LeCun says that within a year to 18 months, we'll have a general method for training hierarchical world models These models would learn from video and real-world data, then help plan actions in robotics, healthcare, and other areas "then scale them toward a universal world model"

@arpit_bhayani Weird times !! Heard one of BA saying programming language will not exist in future. Just listened and walked away

PMs have started shipping code to production, I am just glad that git blame exists.

Most recon tools treat every run as new. We made it stateful. Every exception gets a deterministic fingerprint that survives across runs. February's ₹12,400 variance and March's ₹12,400 surplus collapse into one case. Memory is the architecture, not a feature.

Two ways our reconciliation AI broke recently. Picked a date column as a join key. Emitted gross when it should have been net. Both silent. Match rate looked fine. We stopped measuring AI accuracy. We measure whether the same mistake can ship twice. It can't.

@nikhilkamathcio Had a brief opportunity to work on this with a Swiss client on a technical side, they already have this in another form as gold backed. dgld.ch. Its recently acquired by refinery MKS-PAMP. I believe the Indian Refinery have to step up on this and this can be done.

The world still runs on the dollar. But countries are quietly hedging out: buying gold, trading in non-USD pairs, building payment rails outside SWIFT. UPI has been incredible for India to say the least. To friends championing dollar-backed stablecoins, specifically dollar-backed, this seems like a bad idea long-term for India. Credit where due, to Modi govt and regulators, you got this one right in the face of a lot of pressure. On the other hand, if there were a gold-backed stablecoin and one could monetise the unutilised gold sitting in Indian households to return a yield… don’t know enough to talk about this, but thoughts?

@sabeer Spot on, ecosystem is aligned for builders but yet adapting fast is important, the pace at which things are evolving getting overwhelmed is natural but continuing and being conscious of your pace to adopt things will be more important.

The world’s biggest challenge in the age of AI will not be learning new things - it will be unlearning old ways of thinking and adapting fast enough to a completely new reality.

If you're a CA / GSTP / fractional bookkeeper serving Indian SMB clients: reconpe.com/bookkeepers/ap… (3 months free for the intro cohort) If you're a founder asking whether software is enough: reconpe.com/find-a-bookkee…

ReconPe takes a flat monthly listing subscription from bookkeepers. No per-engagement commission. No escrow. No hidden margin. Engagement money flows directly between you and the bookkeeper. Commission marketplaces incentivise vendor lock-in. We don't.

The Verified badge on the ReconPe directory is granted only after we manually: — Look up ICAI membership — Verify GST practitioner ID on GSTN portal — Phone two client references — Run a 30-min vetting call Never auto-issued when someone enters a number.

Cost math: Hourly rate: ₹500–2,500/hr Retained engagement: 8–20 hrs/month Monthly retainer: typically ₹10k–50k Compare full-time controller: ₹8–15 lakh/year fully loaded. For a ₹1–5 cr ARR business, fractional is the right shape.

The trigger isn't usually dramatic. It's the GST notice with a 30-day reply window. It's the Flipkart variance that needs a dispute filed inside 14 days. It's the TDS mismatch on Form 26AS. None of these are matching problems. All of them need a credentialled human.

The thresholds where founders should hire a fractional bookkeeper: 500+ orders/month 2+ marketplaces Multi-state GST 20+ disputes a month Annual audit cycle Below those: software + founder time often suffices. Above: a fractional saves more than they cost.

Built a reconciliation engine for Indian e-commerce. The most common follow-up question from customers is: "do you know a CA?" Software handles the matching. People handle the judgement. So we built a marketplace for the judgement layer.

Most people think Finance is about math. It’s actually about data cleaning. A typical month-end close doesn't start with a clean ledger. It starts with: - Scanned PDFs - Messy Marketplace CSVs - Bank exports that don't match ERPs - Local-language invoices Before a single transaction can be matched, someone has to spend 3 days just "preparing" the data. That’s not a finance workflow; it's a manual data-entry tax. In 2026, we shouldn't be paying people to be human bridges between spreadsheets. AI should be the bridge, and Finance should be the architect. Agree?

True. It’s the 'Statelessness' trap. Even with Google and ChatGPT, humans tend to treat every financial decision as an isolated event rather than a sequence. Unless the tools we use have a built-in memory to surface the 'why' at the point of purchase, we’ll keep defaulting to the path of least resistance.

When it comes to personal finance, people somehow keep making the same mistakes over and over again. There’s very little creativity in the mistakes people make. Take investing. Pretty much every influencer, every serious finance writer, and the financial media have been screaming for years: don’t mix insurance with investments. ULIPs are usually a bad idea. Endowment policies are usually a bad idea. And yet, ULIP sales continue to grow and endowment plans continue to be sold. People continue to fall for the same pitches, despite all the articles, videos, and excel sheets explaining why these products are bad. The same applies to health insurance, though I have a little more sympathy there. Health insurance is genuinely complicated. There are tiny clauses, room rent caps, waiting periods, exclusions, and conditions that most people don’t fully understand and then they find out the hard way, when they still have to pay out of pocket despite having a policy. But with products like ULIPs and endowment plans, there’s no excuse. These are not impossibly complicated products. Even a cursory Google search will tell you the problem. And today, in 2026, you can just ask ChatGPT or Claude whether a product is a good idea, and they’ll usually show you the math, explain the catches, and give you pointers on what to do. And yet, people still keep falling for the same thing. @PrateekLearnapp from @Zero1ByZerodha has made a really nice video on some of the biggest mistakes Indians make with investing and health insurance. It’s worth watching, and sharing with your friends and family too.

@harshilmathur So true. If you have a niche and a pain point you can address, the ecosystem is perfectly aligned. It really is the Age of Builders.

It’s the best time to build all the ideas you’ve ever had - the app you wished existed, the design you thought was obvious, the library you always wanted. And finally realise why nobody built them :)

@Razorpay We’re officially moving from 'Software as a Service' to 'Software as an Employee.' When the dashboard starts diagnosing and drafting instead of just displaying, the ROI isn't just efficiency—it's the elimination of decision fatigue. Love seeing this shift.

Failed payment = Frustrated customer = An ops headache for every founder Well, not anymore. Razorpay's Agentic Dashboard lets RayAI handle failed payment resolution end-to-end. From diagnosing the issue to generating a payment link to writing the customer email, it does it all. All that's left for you to do, is hit send. #Razorpay #AgenticDashboard