Anenth Guru @anenth

@amanunreal Thanks. This context helps.

Would you agree to this view? In any case, good to know that Opus 5 isn't unbiased.

I agree that Claude is generally a better pair programmer. But in my recent experience, GPT 5.6 Sol was able to tackle a well-defined implementation plan and use sub-agents effectively to delegate. So I was exploring Claude to be a brainstorming and planning system, with Codex as a workhorse.

@_anshulgarg and I teamed up during the hackathon last week to build Sanjay, an assistant for the visually impaired. India has roughly 9 million blind people and tens of millions more with low vision. Bridging the gap in the local language becomes a necessity. Though our inclination was to a use a sovereign stack, powerful vision models are still a work in progress. So we used Gemini. Voice Agent is powered by Sarvam. Checkout the demo!

open.substack.com/pub/mltrenches…

An engaging session from Dhurv Nigam from Dream11 on building a Full duplex voice AI called Googly Bhai. The demo was fun! An interesting observation from the talk. In a chat bot interaction, when the response is immediate, users precieve it as a bot and hence a forced "..." typing state is required. Whereas in a voice agent, even a 1 second delay makes it look a bot and hence a lower latency per turn is non-negotiable. #fifthelephant

keyboard_arrow_left Previous keyboard_arrow_right Next

Good start to the day of volunteering at #fifthelephant conference by @hasgeek

Excited to build a local context aware voice agent at Sarvam epoch buildathon by @GrowthX_Club. Loving the vibe here!

No more guessing when the school bus is actually showing up! Here is the breakdown of the build: The Brain: An ESP32 fetching live JSON data via Wi-Fi. The Data: Pinging the School Bus API for live location & HERE maps for driving distance. The Output: A custom LED circuit that visually alerts us when the bus is getting close!

keyboard_arrow_left Previous keyboard_arrow_right Next

Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available. thinkingmachines.ai/news/introduci… Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵

Something I have been thinking about: in the past, the best engineers I knew spent a lot of time automating their work in various ways. Better vim/emacs automations, writing lint rules to catch repeat code issues, building up a suite of e2e tests so they don't need to smoke test the app manually. These kinds of things were the highest leverage activities an engineer could do, because it multiplied their own output, which in turn meant they could build more things. I think many of these automations have become even more important now. This is true for a number of reasons. First, infra and DevX automation speeds you up. And if you are running an army of agents, each of those agents will be sped up also. More automation == more output per unit of time. Second, moving things to code improves efficiency. Your agent could fix an issue every time it sees that issue happen, but that uses tokens and might miss cases. If Claude instead writes a lint rule, CI step, or routine, that class of issue can be fully automated forever. This is really what people are talking about when they talk about loops -- it's about automating entire types of busywork rather than solving them one off. This isn't a new idea at all. Engineers have been doing this for a long time! Third and most importantly, automation makes it possible for others to contribute to the codebase more easily. Increasingly what I am seeing is engineers are contributing to codebases on day one because Claude can navigate the codebase for them, and that non-engineers are able to contribute to a codebase as effectively as engineers can. What gets in the way of both of these is domain knowledge that lives in peoples' heads rather than in automation -- the stuff you used to have to learn when ramping up. What has changed thanks to agents is the domain knowledge that can be encoded as infrastructure is no longer limited to what is expressible in lint rules and types and tests; it can now capture nearly all domain knowledge, encoded as code comments and skills and CLAUDE.md rules and memories. If I put up a PR for an iOS codebase I don't know and a code reviewer rejects it because it doesn't use the right framework, or if a designer builds a new feature and it gets rejected because it doesn't follow the right architectural patterns, these are failures of automation. Every team should be writing the CLAUDE.md's, REVIEW.md's, skills, and docs that enable agents to productively work in their codebase with zero additional context from the prompter. This sounds crazy, and at the same time is a natural extension of the stuff engineers have always done: automate, and encode domain knowledge as infrastructure. As the model gets smarter and as the harness matures, this task becomes easier. In the meantime, it is on every team to look for ways to convert their domain knowledge to infra so that Claude can write code better, so that code review catches issues automatically, and so the next person working on your codebase can contribute more easily.

I believe you were referring to plateau in the demand. I was referring to the upcoming trough in the devs headcount. I agree that demand will continue to grow. As agents become more and verify their own work end-to-end, 'hint at the feature, get the software' becomes real. At that point, the leaders gain the confidence with the newer system, and it would trigger the workforce trim.

@CharlesUchi_Dev 😀

Built a custom motion-capture gaming rig without buying any new hardware. 💻🚲 I prompted Claude to write a computer vision script that turns a standard laptop webcam into a precise motion tracker. It reads the physical handlebar positions and pipes it straight into the motorcycle game on the big screen as a 1:1 controller input. Watch the sync between the physical turns and the gameplay. ​ #BuildInPublic #ClaudeFable #ComputerVision #NodeJS #IndieDev #GenerativeAI

@victoressienDev All the best. With a coding assistant, all the hard parts for a novice is gone. I never imagined to build something like this a couple years ago.

I have been having a lot of fun lately with this hobby hardware project! #arduino #esp32

keyboard_arrow_left Previous keyboard_arrow_right Next

Keen to collaborate? Please DM.

Reading dense technical papers is painful. At the recent @AIBoomi hackathon, @arvind_devaraj & I built jusRead. It uses LLMs to pick the right format (audio, video, interactive widgets) for each section and makes the learning experience engaging. Demo below: a paper on how LLMs work under the hood.

youtu.be/xyCOvT1Y5YQ?si… This 19-year-old interview of @elonmusk has aged well. Inspiring to see his conviction on the long-term bets and how well they turned out for him.