Ankit Daftery

@heydambis @nileshtrivedi ^^

Have a cute story behind becoming a broker now but hey there's a 3bhk available for rent at 9th cross, 12th Main Road Indiranagar, right behind the fresh factory. 50k rent, 1.5L deposit. 1 car parking space. DM for deets, RT for reach.

When the question is ready, the answer will appear

Seems like a logical sounding but incorrect metric to create demand. You can rewrite your entire code base every couple of days by consuming tokens, be no closer to doing anything that makes the product or business better. I expected better.

@unitechy @WisprFlow The immediate bottleneck you hit after choosing the model is context quality and length. Humans are pretty lazy at giving long form context so it makes sense to solve laziness by replacing typing with speech. Lots more opportunities are still open here IMO

@ankitdaf @WisprFlow If you go my replies, looks like everyone is solving for this now suddenly?

This @WisprFlow is becoming worse day by day 😞 Had to type this post

@shantanugoel I built it out as a local daemon and then came across it, it's well packaged. I added Openvino IR supported models and GPU offloading for better performance and it worked well. On my wishlist is a post processing step for spell check and rewriting format (list, email etc)

@ankitdaf I saw that only half an hour ago so haven't checked it out. My inspiration was mainly wispr flow + a long personal wish list 😄

Started building my own Wispr Flow alternative - Aawaaz (sound/voice) Going to be Free, Local-first (or local-only), Privacy friendly, yet super fast and super accurate! Basic transcription working well already and will make this a top-class alternative very soon!!

Looks like @HDFC_Bank has broken SI Hub . There's a username and password field but no button to click after entering those

@tapedspectacles @unitechy @narendramodi @upliance cancelled in one tweet, brutal ! 🤣

@ankitdaf @unitechy @narendramodi @upliance When I see Ads 😛

are we supposed to buy induction cooktop @narendramodi ji pls to guide 🙏🙏

@_svs_ @sameeracan How are GPUs not in the mix already as joining bonuses? It's 2026 ;)

@sameeracan Do you send them laddus when they accept the offer?

10 years in recruiting and I can barely predict last minute candidate back out AI huttttttttt

@priteshlakhani You can buy something similar on Chinese websites, not sure if this needs a specific kind of phone and software, will look into it tonight

Can someone explain how to implement this?

@ku1deep Are you going to drop a rebuttal or just leave everyone hanging?

Look if you are LARPing as AI expert please don’t react to that “something is happening” article. You will give yourself away.

@_swanand @championswimmer Heh, while true, I believe the problem has now shifted to *which* plan dot md to spend your mental bandwidth on more than jumping to code directly

@ankitdaf @championswimmer Then: Weeks of working can save you hours of planning Now: Millions of tokens on code can save you thousands of tokens of plan dot md

Give half the tasks to Claude, half to Codex, wait for both to raise their PRs, merge them, cut a release. Insane speed of development, and I can understand why people doing this are starting to go a bit crazy - hard for brain to keep up with this speed.

@bhavintu Humans are still on average much better at "jumping to the end", figuring where computing tokens are unnecessary. This adds a huge multiplier to the "value" of each token a human does output, which might be a better metric to measure than just the "count" of tokens

Cost of Humans vs Cost of AI - Human * Avg rate of Speech output: 150 words per minute * Avg wages per hour for skilled worker: $50 per hour * Avg cost per million spoken words: $5,555 AI cost per million tokens * GPT-5.2: $14 * GPT-5-mini: $2 * Opus 4.5: $15 Comparison Notes * Avg humans are 396x more expensive than sophisticated models * The above analysis is quite inaccurate and biased towards humans due to the following factors * Humans tend to take more time to think and so actual output per minute is not close to 150 words * If we take typing instead of speech then human output per minute reduces further * If we then count average wages of lawyers / scientists / senior engineers / knowledge workers the cost disparity further increases * Above and more factors likely introduce a further 10x efficiency * In reality it seems that AI is likely 1/4000th the cost of a human for outputting the exact same number of tokens as a human for a given problem

@abhijitwt You should be happy, for the first time there's a "developer quota" somewhere ;)

This is the Rajasthan government’s official website for bus ticket booking. You will not receive any OTP on your phone or email. To get the OTP, you have to open the browser’s Network tab and inspect the request manually. In short, only a developer can book tickets on this website I don’t know whether I should laugh or cry?

@regalstreak @Swiggy How much token cost (in INR) do you incur per order ?

connected @Swiggy to clawdbot and automated my biggest headache! now it knows my schedule - chai delivered to my door at 5pm - auto builds carts and orders groceries for meals i cook - dinner recommendations based on what I'm craving setup takes ~ 1 minute just ask clawdbot to set up the swiggy skill I've published, everything else runs on autopilot!

My only advice to CEOs this year.. Hire tinkerers. Make it high status. Find ones that will explore the edges and (ideally) naturally gifted at teaching people. Empower them to freely roam across the org and fix large problems that can be automated. All your execs will complain that these tinkerers don’t understand scale or systems (rollout being a favorite word). Listen, and ignore them. Or even better give them budget to hire a tinkerer to achieve their targets. Give these tinkerers ambition, purpose and hard targets and watch them fly.

@surbhiskjain Quinoa salad does wonders for me here.

I have lunch at 1 pm and dinner around 7–7:30. But I’m always hungry at 4:30/5 pm. What’s a healthy, filling snack (not too heavy) that can carry me for 2–2.5 hours?