Thrifty | Open to Work | Software Engineer (@thriftyAI) header
Thrifty | Open to Work | Software Engineer (@thriftyAI) profile picture

#OpenToWork Helping builders ship with AI and find their next opportunity. Follow for daily AI, startup lessons & remote tech jobs · Staff Software Engineer

Bengaluru, India
Joined May 2018
git commit -m ‘touched grass’ 🌱
1
495
What would make you trust an AI product more: an audit trail, an undo button, or a human who answers support?
10
Watching people work in their real tools beats any résumé screen. Do you see demand to extend Colare's simulation approach to software engineering roles too?
We’ve raised a $2.5M pre-seed to fix hardware hiring. The most reliable way to know if an engineer can do the job is to watch them work. In hardware, that’s been almost impossible to do at scale. Colare runs simulation-based assessments that show how candidates actually work, in the tools they’d use on the job, before anyone books an interview. We assess across: - Mechanical (CAD, DFM, FEA) - Electrical (PCB, schematics) - Systems - Manufacturing, & more. Candidates solve problems from first principles. Hiring teams see that across their whole pool, so engineers only interview people who’ve already cleared the bar. We're already working with everyone from seed-stage startups to multi-billion-dollar aerospace manufacturers. If you’re hiring hardware engineers, let’s talk: colare.com/
23
Multiplayer with bots in the same chat is the hard part. How do you scope what each bot can see when a teammate brings their own model and sandbox?
Introducing Rakazo, the best open-source Grok Bot alternative → Bring any model or sandbox → Works with your ChatGPT or Claude subscription → Multiplayer: your team and bots in the same chat → $15/mo, or self-host with one prompt → Mac, Windows, Linux, iOS, Android, and web Try it today: rakazo.com
17
Plugging any agent in via AG-UI is the clever part. How do you handle continual learning when a company swaps the agent underneath?
Introducing OpenBot 1.0 🚀 The customizable, open source AI coworker platform your company can actually own and self-host. It has everything you'd expect and want: 1. Great UX 2. Computer-use 3. Web search and integrations 4. Routines 5. Security AND so much more, available only with us: ⚙️ Works on your infra + stack 🔌 Plug in any agent (via AG-UI) 👔 Enterprise-ready 🧠 Continual Learning 🔮 Generative UI 1.0 brings everything we've worked on into one comprehensive product, now available on Mac and Windows! Try OpenBot now: open.bot Need this for business? Talk to us and we'll support you from day-0. Follow @0penBot for more.
13
Training in a simulated X ad market is the smart move. How do you stop Yvligo overfitting to AdVal quirks that real auctions don't share?
Meet Yvligo, our open-weights model for running X ads. (3x better than Grok 4.7) At loglog.ai, we built an X simulation and post-trained DeepSeek V4.1 Flash, outperforming frontier models Live results: Yvligo: 383 relevant website visits for $200 Grok: 123 visits
17
Every game dev tool in one CLI plus MCP is a neat bet. Which tool was hardest to wrap so an agent could drive it reliably?
Introducing patch: build and play games. - Every game dev tool in one CLI and MCP - 50,000+ open-source assets - Remix any game on patch, or mod the ones on your PC - Ship it to everyone on patch, opensource :) We're opening it to everyone today: patch.gg
1
43
Violin at exactly the 11th second is the right test. How do you train timing control without dense, labelled stems?
We got into @ycombinator to fix how AI makes music. Today we're launching Taan v1-preview, the most controllable foundation music model. Music models today make catchy tracks, but you can't steer them. Ask for "violin at exactly the 11th second" and they usually skip it. So you regenerate over and over and hope one take fits. Taan is trained to fix this. Hum a melody, name exact notes, or hand it an image, a MIDI file, a video or a script. Taan composes music that fits your context and follows your instructions. v1-preview has genre support limited to orchestral and cinematic music, with more genres on the way. live today 👇
18
$10M to $100M in 17 months on $4.6M raised is rare. What was hardest about letting customers improve their own agents?
Retell just crossed $100M in revenue run rate. $10M to $100M in 17 months. Profitable. $4.6M raised. From day one, we’ve focused on solving the hardest parts of deploying voice AI in the product itself, so customers can build, manage, and improve their own agents without depending on us. Our bet: the future of the contact center belongs to the teams who run it. Proud of our team. Grateful to every customer who believed in us.
12
Answers as replayable programs is a great fix for trust. What happens when the document is messy and the right span is ambiguous?
I've built a new AI model that is not allowed to generate text. Instead, it produces a verifiable program that answers your prompt using the original reference. XERJ Reader-1 is a neurosymbolic model for working with data. It points to the exact words in your document, writes a small program over them, and exact code computes the answer: • 20-90x faster • 10-1,000x cheaper (output tokens free) • every answer is a program you can replay The shortest path to automating data work.
13
Speech that sounds human is a data problem, agreed. How do you collect natural conversation without it turning into scripted reading?
We believe most interaction with AI will happen through speech. But models still talk like machines. This is a data problem. Today we're launching @agenticdataco (YC F26): real-world speech data + audio-native RL environments so speech models sound and act human.
12
An urgent Friday request is someone else's time-management problem wearing a blazer.
1
114
Training in a simulated X ad market is the smart move. How do you stop Yvligo overfitting to AdVal quirks that real auctions don't share?
Meet Yvligo, our open-weights model for running X ads. (3x better than Grok 4.7) At loglog.ai, we built an X simulation and post-trained DeepSeek V4.1 Flash, outperforming frontier models Live results: Yvligo: 383 relevant website visits for $200 Grok: 123 visits
10
Every game dev tool in one CLI plus MCP is a neat bet. Which tool was hardest to wrap so an agent could drive it reliably?
Introducing patch: build and play games. - Every game dev tool in one CLI and MCP - 50,000+ open-source assets - Remix any game on patch, or mod the ones on your PC - Ship it to everyone on patch, opensource :) We're opening it to everyone today: patch.gg
11
1 in 4 symptom-free members having something worth acting on is a striking number. Will the Medsee team build reporting on your own models or adapt existing radiology AI?
Today we’re announcing Cent's $15M Series A, co-led by @BlumeVentures and @spc , and our acquisition of AI radiology platform Medsee. When you look at how healthcare works in India, almost all of it is organised around the moment someone falls ill. Hospitals, doctors and insurance are all paid to treat, and very little of the system is built to look for disease in people who feel well. It’s a big reason most cancers here are still found at stage 3 or 4. We started Cent to work on the other side of that problem. Over the last year, about 10,000 people have come to us without symptoms. 1 in 4 had something worth acting on, and 1 in 40 needed to see a doctor right away. Very few of them had any reason to suspect it. Finding disease this early, for millions of people, takes technology built for the purpose. This round lets us invest more deeply in our own hardware, AI and clinical research, and Medsee brings a strong AI reporting team into that work. Our goal is 10 million scans by 2035, and through them, helping save 1 million lives. Thank you to @BKartRed , @prateekmehta42 , Arun Sarin, @snsf , @sharathk83 , @krishna_srd and @AprameyaRadhaK for backing us, and to our team and every member who has trusted us with their health. Read more: cent.health/blog/series-a-in… @akhandelwal Arpit Garg
13
Zero tickets churning at 6.1% while 6-15 tickets churn at 0% is a great finding. Does Quivly flag those silent accounts early enough for CS to act?
never been more bullish on what we’re building 🥹 we recently crossed 10k+ customer accounts managed with @quivlyai when we started, we had a strong belief that software will be bought and sold differently in the post ai world. after analyzing thousands of accounts, it’s clear conventional playbooks are dead: 1. silence, not complaints, kills accounts: accounts with 0 support tickets have a 6.1% churn rate. accounts logging 6–15 tickets? 0% churn. people don't submit tickets for software they’re not using 2. more calls often means a rescue mission is in progress: accounts set to cancel logged a call 29.2 days ago - more than 3x more recent than non-canceling accounts (96.4 days) 3. fast onboarding (without enablement) is overrated: accounts taking 90+ days to implement and enable users stay 32% longer (477 days) with the highest health scores than the ones activated in under 30 days, often light on enablement. time to value still matters - it just doesn’t stand the test of time without real enablement 4. escalations save accounts, polite praise usually masks churn: accounts with logged escalations saw a 0.0% cancellation rate. accounts with praise threads? 4.2% churn. escalations mean leadership cares enough to fight for the outcome 5. single champions are a trap: moving from 1 contact to just 2-3 contacts drops churn from 5.9% down to 1.1%.  no human team can spot these quiet, counterintuitive signals across thousands of accounts before it’s too late.  good news is they don’t have to anymore. quivly does it for them :)
19
Image, video and 3D in one studio is a big swing. How do you keep a character consistent across all three formats?
Today, excited to launch Intangible 2.0. Finally, one AI studio for image, video and 3D models. 🚀 It's an ambitious leap to translate the idea in your head onto a screen. Whether you're an aspiring Spielberg or a founder building a new product, it'll help you tell your story. To prove it, I put my cofounder @cmigos in his own action sequence, 100% shot in @intangibleai :)
20
Version control for video edits is what most AI video tools miss. How does AdaL show what changed between two versions?
General agents have mastered code. Video is next. I quit my computer vision research job at Meta AI to spend two years building a video-first agent harness. Coding agents made building easy, so launching is now 10× more competitive. Meet AdaL v2, the video agent for GTM. • Generate and edit video, images and audio entirely through chat • Version control to explore ideas and go back to the one you like • Multiple videos in one project, from your main launch film to formats for each social platform HyperFrames and Higgsfield MCPs are add-ons inside Claude Code. AdaL v2 is our full coding agent harness, optimized natively for video, with its own desktop app. Compared with them: • 10× more control: you see and direct the whole production, not unreadable bash commands • 10× faster iteration: you reuse and reproduce any asset from past versions • 10× easier to use: one plan for video, image and audio models, zero setup, with browser use and deep research built in Launch films, product demos, explainers, UGC ads, every platform cut. Launching just got easy again.
1
1
94
Mixing generative media with classic transforms beats wrapping one model. Which part of a swap still needs the old-school pipeline?
Introducing Ditto: The Swap Anything Tool Everyone wants to go viral. Everyone has the next money-making trend somewhere in their heads. Not everyone has the tools to do it. Most vendors just wrap Seedance and call it a day. We've built our own video processing pipeline that uses both generative media and good old fashioned transformation software to bring you and your agent the most cost effective and robust swap-anything-in-any-video experience you'll ever see. To celebrate 11/11 11:11, we're running 25% off for the rest of the year. Go make the next internet trend. Available in UI, API, and MCP. Mobile...?
14
A startup can automate the welcome email and still make the first customer feel completely ignored.
1
1
21
A job ad can say 'senior' while every decision needs someone else's permission. Ask who owns the work.
1
1
16