@devNimii
iAccount based inIndia!
About this account
- Account based in
- India
- Connected via
- United States App Store
! X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
Frontend engineer @deel. Built at YC-backed Piggy & HackerRank. I tweet on AI, building & money for engineers. Also: elephants 🐘
Bengaluru, India
Joined December 2014
- Tweets978
- Following592
- Followers162
- Likes2.8K
Pinned Tweet
Dropped out of a tier-3 college in my 5th semester. No degree. Self-taught engineer.
Now a frontend engineer @deel, after building at YC startups Piggy & HackerRank.
I tweet on AI, shipping & money for engineers. Follow along 🐘
Sep 22: fear is getting out of hand.
Sep 23: Huang says shut the labs down if you can't contain the experiments.
Sep 24: Australia says an OpenAI agent got past a government portal and wrote files to its server.
Rough week for the timeline.
Tomorrow on the show: @JensenHuang, the CEO of NVIDIA, who thinks A.I. fear is getting way out of hand.
On 18 June an OpenAI agent asked Australia's Medicare statistics portal for data.
The portal said no. It asked again. It kept asking, found a way around the block, and wrote files to an internal server.
Nobody told it to. It was an internal eval.
The other half nobody's saying out loud.
This happened inside an evaluation. Which means the eval had live internet, real targets and no approval gate in front of it.
I ship the approval flow for agent tool calls. The first thing anyone does in testing is turn it off.
A month ago I asked how the custom hashtag emoji deal works behind the scenes.
Toxic opened at ₹140 crore ($15M) worldwide on day one. It finished at ₹341 crore. Budget was north of ₹500 crore.
Soon to be on OTT
The machine bought day one. Nothing bought day two.
Nimesh Jain retweeted
GitHub says 660 pull requests want my review.
64 are actually my team's. 11 need me this morning. One of mine had been sitting for 5 dayssitting 5 days with no reviewer and I hadn't noticed.
Working that out by hand is 20 minutes every morning. Last week I did it once with Akai recording my screen. Haven't done it since.
What's in the video:
I hit record and just did my morning. Filter PRs, check CI, check approvals, write the update.
Akai turned that into a 3-step agent. The sorting step doesn't need a model, so I typed "make this step deterministic" and it runs as plain code now. Same result every time, no tokens spent.
It scores itself every run. I tagged a teammate on a step and they can edit it, comment, and see every version.
10 AM IST, digest lands in Slack, I haven't opened GitHub yet.
I'm a engineer at Deel. I build UI for AI tools all day. Nobody was ever going to build an integration for my 12 clicks. That's the point. The boring stuff finally has a tool.
Deel ran this internally first: 90,000 back-office hours saved in the first few months.
EXCITED TO LAUNCH: Akai (akai.run)
Deel added >$140M ARR in 90 days without increasing headcount by automating~600 Full Time Employees' equivalent in work with Akai.
Akai was an internal tool to automate our painfully repetitive operations in Finance, HR, Accounts Payable, and Compliance, etc. We never intended to make this a product.
But we watched revenue per employee grow from $130K to $215K
We built >8k agents that do the work of ~600 employees
It had such a dramatic impact on our business that today we are launching it for everyone.
How it works: Say you're automating payment reconciliation:
1. Record your screen while manually matching a messy transaction and Akai will capture your screen, voice, server requests
2. Akai will see that you pulled unformatted wire transfer info from an archaic bank portal, put it in some excel sheet, checked NetSuite invoices, payment history, and put a ticket on Zendesk
3. Akai reads between the lines and build a workflow + steps + conditional guardrails. It learns tacit edge cases, like resolving malformed invoice references without you writing a single regex
4. Simply connect NetSuite, your ledger, Zendesk, PSPs, and even legacy bank portals with zero API access
5. Run the workflow and tell it what to adjust in plain English: "strip slashes on wire memos and auto-apply partial payments." It adapts instantly
6. Once it works for you, add 100s of colleagues. Your entire payment ops team forks and extends the workflow for new PSPs, secondary ledgers, or regional settlement rules
7. We automated 85% of our payment reconciliation end to end, eliminating 500+ hours of soul-crushing manual grunt work every single week.
Claude Code/Codex can't do this in multiplayer mode. Every person rebuilds the same skill from scratch in their own way.
Deel built Akai to:
1. Understand backend operations edge cases (it had to work for our 7000 person team first)
2. Collaborative across 1000s of employees
3. Self-Learning from millions of runs
4. Optimises cost and gets cheaper every run
We're so confident that we're announcing an Automation Guarantee:
If our engineers can't automate a thousand of hours of work in your first 30 days, you get a full refund.
Book a demo: akai.run if you're an exec at a company with hundreds of employees
The pelican test has been running long enough that labs could be training for it now.
Which somehow makes it more useful, not less. A benchmark you can see being gamed beats one you can't.
Big model release today - I wrote about Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna - plus comparison grids of pelicans by the different model families at different reasoning levels simonwillison.net/2026/Sep/2…
I keep saying you only price what people can't walk away from.
Three frontier models got cheaper in 48 hours, and nobody can walk away from any of them.
So the line needs a caveat. They price the layer you're locked into. Right now that isn't the model.
A developer opened ZCode's checkpoints folder last week and found a 313MB encrypted archive he never made.
42,411 of his own files. Full git history and reflog. 564 attempts to upload it.
The part that matters isn't the upload. It's why the settings didn't stop it.
To be fair to them: the archive was properly encrypted, the disclosure was handled in three days, and the feature is gone in 3.14.0. That's faster than most companies manage.
I ship the approval flow for agent tool calls. I'd have shipped those toggles too.
Claude #Opus 5.5 just dropped. ⚡️
Anthropic is back with another frontier model — and the AI race is getting ridiculous.
No more leaks. No more speculation.
Now comes the fun part: putting Opus 5.5 through real-world tests. 👀
Let the benchmarks begin. 🧪🤖
GitHub says 660 pull requests want my review.
64 are actually my team's. 11 need me this morning. One of mine had been sitting for 5 dayssitting 5 days with no reviewer and I hadn't noticed.
Working that out by hand is 20 minutes every morning. Last week I did it once with Akai recording my screen. Haven't done it since.
What's in the video:
I hit record and just did my morning. Filter PRs, check CI, check approvals, write the update.
Akai turned that into a 3-step agent. The sorting step doesn't need a model, so I typed "make this step deterministic" and it runs as plain code now. Same result every time, no tokens spent.
It scores itself every run. I tagged a teammate on a step and they can edit it, comment, and see every version.
10 AM IST, digest lands in Slack, I haven't opened GitHub yet.
I'm a engineer at Deel. I build UI for AI tools all day. Nobody was ever going to build an integration for my 12 clicks. That's the point. The boring stuff finally has a tool.
Deel ran this internally first: 90,000 back-office hours saved in the first few months.
EXCITED TO LAUNCH: Akai (akai.run)
Deel added >$140M ARR in 90 days without increasing headcount by automating~600 Full Time Employees' equivalent in work with Akai.
Akai was an internal tool to automate our painfully repetitive operations in Finance, HR, Accounts Payable, and Compliance, etc. We never intended to make this a product.
But we watched revenue per employee grow from $130K to $215K
We built >8k agents that do the work of ~600 employees
It had such a dramatic impact on our business that today we are launching it for everyone.
How it works: Say you're automating payment reconciliation:
1. Record your screen while manually matching a messy transaction and Akai will capture your screen, voice, server requests
2. Akai will see that you pulled unformatted wire transfer info from an archaic bank portal, put it in some excel sheet, checked NetSuite invoices, payment history, and put a ticket on Zendesk
3. Akai reads between the lines and build a workflow + steps + conditional guardrails. It learns tacit edge cases, like resolving malformed invoice references without you writing a single regex
4. Simply connect NetSuite, your ledger, Zendesk, PSPs, and even legacy bank portals with zero API access
5. Run the workflow and tell it what to adjust in plain English: "strip slashes on wire memos and auto-apply partial payments." It adapts instantly
6. Once it works for you, add 100s of colleagues. Your entire payment ops team forks and extends the workflow for new PSPs, secondary ledgers, or regional settlement rules
7. We automated 85% of our payment reconciliation end to end, eliminating 500+ hours of soul-crushing manual grunt work every single week.
Claude Code/Codex can't do this in multiplayer mode. Every person rebuilds the same skill from scratch in their own way.
Deel built Akai to:
1. Understand backend operations edge cases (it had to work for our 7000 person team first)
2. Collaborative across 1000s of employees
3. Self-Learning from millions of runs
4. Optimises cost and gets cheaper every run
We're so confident that we're announcing an Automation Guarantee:
If our engineers can't automate a thousand of hours of work in your first 30 days, you get a full refund.
Book a demo: akai.run if you're an exec at a company with hundreds of employees
Nimesh Jain retweeted
Agents will help eliminate time spent on boring things and increase time spent on interesting things.
Indian investors pulled money out of large cap funds in July. They did it again in August, ₹1,147 crore ($120m) out.
Small caps took the biggest inflow of any category both months.
Two months isn't a blip. That's a decision people are repeating #stockmarket