Thinking in Public

A personal archive for independent long-term builders — engineering, A.I., drones, systems, and the life around the work.

Engineering & Life

Latest Articles

How Kokoro Uses Dilated Convolution

How Kokoro Uses Dilated Convolution

Kernel math / 03A three-tap filter can listen farther in time without gaining another weight. Kokoro uses that idea, but the filter is only one part of making speech. In the first two parts of this series I took convolution from wx+b to a moving window, then followed its shared weights backward through a gradient. Now I want to see what happens when the window does not touch every adjacent sample...

Subscribe

Get my rants delivered to your inbox

I will send new posts as and when I write. No fixed cadence, just engineering notes, rants, and things I am thinking through.

Micro-Log

Quick Updates

Need an intelligent system to work on real hardware?

Embedded systems · Robotics · Constrained AI · CPU and HPC · Accelerators · Distributed systems

Work With Us / Antshiv Robotics

Engineering & Life

Articles

Backpropagation Through Convolution: Shared Weights, Overlapping Gradients

Backpropagation Through Convolution: Shared Weights, Overlapping Gradients

Kernel math / 02In part one, one filter moved across a signal. Here we make that filter learn: one forward pass, one loss, every backward contribution, and one measured parameter update. Before differentiating anything, I want to see the actual shapes. Imagine a tiny grayscale image row with five pixels. We move a three-number filter across it, one pixel at a time. The filter produces three outpu...
Convolution Is Wx+b Moving Across a Signal

Convolution Is Wx+b Moving Across a Signal

Kernel math / 01 From wx+b and GEMM to the sliding filters that help modern models see and listen. This continues my series on wx+b and backpropagation and matrix multiplication, or GEMM. I want to look at another piece of older mathematics that remains critical in AI: convolution. It was useful in signal processing and image analysis long before today's generative models. Now local filters...
From Generated C to Useful Agents: CKE Serving Qwen Code and Gemma4

From Generated C to Useful Agents: CKE Serving Qwen Code and Gemma4

A model giving me text is useful, but it is not yet an agent backend. An agent needs to send a conversation, receive a structured tool call, run that tool, return the result, and keep going. If a request is cancelled or a tool fails, the next request still needs to work. CKE has now crossed that boundary in two local CPU workflows. Qwen Code completed bounded file-editing tasks against a Qwen...
CKE Vision: Find the First Bad Edge

CKE Vision: Find the First Bad Edge

A vision model can return plausible text while taking the wrong path through its image prefix. That is the trap I want CKE's tests to catch. In October, the Gemma4 and multimodal work was less about adding a shiny new model name and more about making the first image-to-decoder boundary inspectable. The September CKE recap described three streams: model architecture, modality bring-up, and str...
CKE's First Kokoro WAV Is Not Ready Yet

CKE's First Kokoro WAV Is Not Ready Yet

cke cpu ai kokoro tts 4 min read
I want my Linux workstation to read an agent's answer aloud. Whisper already helps me transcribe my own speech; Kokoro is the other direction. I have been following its path through CKE one generated boundary at a time, from phoneme embeddings to duration, prosody and finally waveform generation. There is now a new milestone: a fixed Kokoro request produces a 61,800-sample native waveform and...
What Two Requests Actually Mean for CKE Serving

What Two Requests Actually Mean for CKE Serving

When I say I want to run several models across my own machines, it is easy to jump straight to a diagram with six GPUs or a cluster of CPUs. The harder question is what one user request actually occupies: model weights, KV cache, recurrent state, a decode slot, and a place in the queue. Those are different resources. CKE's October serving work made that distinction concrete. The CKE serving a...
What Changed in C-Kernel-Engine This September

What Changed in C-Kernel-Engine This September

September was the month C-Kernel-Engine started to feel less like a collection of model demonstrations and more like a system I could keep extending. That is not the same as saying it is finished. I spent much of the month finding places where a model appeared to work until a real recording, a real tool call, or an independent numerical comparison asked a harder question. I wanted CKE to run mode...
Kokoro Is Reaching Generated C One Boundary at a Time

Kokoro Is Reaching Generated C One Boundary at a Time

cke cpu ai kokoro tts 11 min read
Kokoro's phoneme encoder just took its first concrete step into CKE's generated-C lane. The step is deliberately small: one operation three embedding tables plus a per-token LayerNorm now lowers through the normal v8 pipeline into generated C, binds five real weight payloads, and matches a pinned PyTorch capture within 2.384185791015625e-7 maximum absolute error, at a fixed 36-token exten
How Kokoro TTS Works, and Where CKE Stands

How Kokoro TTS Works, and Where CKE Stands

cke cpu ai kokoro tts 14 min read
I have been working on speech recognition in C-Kernel-Engine for a while. Whisper can transcribe my own videos. The other direction is becoming more interesting to me now: I want my Linux workstation to read a useful response from a coding agent without making me stare at another terminal. That is why I started looking closely at Kokoro, a small text-to-speech model, and asking what it would take ...
I Want To Use My Linux Workstation Without Looking At It

I Want To Use My Linux Workstation Without Looking At It

I Want To Use My Linux Workstation Without Looking At It I have spent a lot of time building software that lets me run AI models on CPUs. Now I want to use some of that work for a more personal reason: my eyes hurt after too much time on the screen. I still want to work with my coding agents, follow experiments on my CPU nodes, and move between CKE, Antsand, and robotics work. I just do not want t...
From Native Kernels to Generated Audio Circuits: Parakeet and Cohere in CKE

From Native Kernels to Generated Audio Circuits: Parakeet and Cohere in CKE

A few weeks ago I wrote about running Parakeet and Cohere Transcribe across my complete 42-minute recording. That work was real and useful. CKE converted the weights, ran native kernels, produced complete transcripts, retained token trajectories, and compared results across Intel and AMD CPUs. But Python still knew too much about how those models were assembled and executed. That was the next arc...
How Ericsson Edge Routers Shaped CKE

How Ericsson Edge Routers Shaped CKE

A lot of how I design software today comes from the five years I spent at Ericsson working on packet classification for SE edge routers and SSR Smart Services Routers. I did not fully connect that work to C-Kernel-Engine (CKE) when I started CKE. But as its DSL, circuits, kernel maps, providers, and generated runtimes have matured, the influence has become quite obvious to me. If CKE is new to you...
How to Get Started With C-Kernel-Engine on Linux

How to Get Started With C-Kernel-Engine on Linux

If you want to run an artificial-intelligence model on a CPU and inspect the generated program rather than hiding it behind a large framework, this guide shows how to get started with C-Kernel-Engine (CKE) on Linux. We will clone the source, prepare the development environment, download a small Gemma model, generate a C runtime, run one prompt, and open the intermediate-representation visualizer. ...
Beyond Whisper: CKE Runs Parakeet and Cohere Across 42 Minutes on CPUs

Beyond Whisper: CKE Runs Parakeet and Cohere Across 42 Minutes on CPUs

CKE's audio work is beginning to expand beyond Whisper. NVIDIA's Parakeet is now transcribing my complete presentation on an Intel central processing unit (CPU) with no graphics processing unit (GPU), and Cohere Transcribe is doing the same through a different FastConformer circuit. That is useful, but it is also where I need to be careful not to turn successful fixtures into a broad audio-platfor...
What CKE Fixed Today: Honest Evidence and Faster CPU Whisper

What CKE Fixed Today: Honest Evidence and Faster CPU Whisper

CKE made meaningful progress today. Not because another model name appeared in a compatibility table, but because the evidence became harder to misread and several measured CPU bottlenecks became smaller. The work covered three connected problems. First, a nightly report needed to prove which capabilities actually ran on the current commit. Second, a Qwen3-VL failure turned out to be a validation...
How CKE Brought Up Qwen3.8 Flash Next

How CKE Brought Up Qwen3.8 Flash Next

CKE can now run the tested Q4_K_M text artifact for Qwen3.8-Flash-Next. The Q4_K_M quantization format is not experimental. What remains experimental is CKE's support boundary for this particular model: short-prompt Q4_K_M text execution is working and numerically certified, while full-model BF16 parity, multimodal execution, consumed long contexts, and matched performance remain open. This model...
Qwen3.8 Broke In CKE. Our Regression Tests Needed To Catch It.

Qwen3.8 Broke In CKE. Our Regression Tests Needed To Catch It.

I have been writing about how quickly C-Kernel-Engine is adding model families. Qwen3.8 had already produced useful output in CKE, so I treated that support as progress we had earned. Then issue #454 showed a real Qwen3.8 path producing gibberish. A compiler/provider-selection defect had reached a user, and our tests had not stopped it. This was not merely a new user having difficulty. They expo...
I Used CKE To Transcribe My Video. It Found A Bug In My Compiler.

I Used CKE To Transcribe My Video. It Found A Bug In My Compiler.

Yesterday I recorded the second video in my kernel engineering series. Since C-Kernel-Engine already had Whisper support, I wanted to use it to transcribe my own recording. Not another small demonstration. My actual microphone, my explanations, my pauses, and a video I wanted to publish. That should have been a useful small win. Instead, one of the broken runs gave us essentially one word: Th

How Audio Becomes Words: Whisper's Kernels, Step By Step, On A CPU

Speech recognition is mostly signal processing wearing a transformer hat. C-Kernel-Engine (CKE) runs Whisper Tiny, Base, and Small end to end in generated FP32 C — no PyTorch at inference. Because every stage is a named kernel with a parity oracle, the whole path from a WAV file to a transcript is unusually easy to read. This post walks it kernel by kernel, using CKE’s Audio Ker...
Today I Made My Drone Simulation Take Off From A Gamepad

Today I Made My Drone Simulation Take Off From A Gamepad

A working control loop should be understandable before it touches a motor Today I wanted to get back to the Antshiv flight-controller work. I already had the aircraft model, the Logitech controller, the PID and mixer, and the RK4 plant. On paper, that sounds like the joystick should just move the drone. It did not. Sometimes a button highlighted, sometimes the aircraft moved, and sometimes ...
CKE Now Runs More Than Qwen: Cohere, Laguna, Instella, Nemotron And More

CKE Now Runs More Than Qwen: Cohere, Laguna, Instella, Nemotron And More

CKE is slowly becoming easier to extend C-Kernel-Engine (CKE) is advancing pretty well. I now have more compute to parallelize the work across the P3 and Ryzen nodes, and that has made it possible to add more coverage across very different model families. There is still a lot of work left, but the architecture is holding up decently so far. Much of this work is being done with help from A...
One Ryzen, 128K Context, And A 27B Model: What Qwen3.8 Generated With CKE

One Ryzen, 128K Context, And A 27B Model: What Qwen3.8 Generated With CKE

Qwen3.8 is punching way above its weight class When I shared some of the infographics Qwen3.8 generated on my Ryzen CPU beside comparable work produced through Anthropic's Opus, nobody in the room could reliably tell which system made which artifact. My roughly CAD $800 processor was not matching a frontier datacenter on speed or infrastructure scale. But for the general technical reasoning...
CKE Added Qwen3.8 In About An Hour. The Architecture Is Starting To Pay Off

CKE Added Qwen3.8 In About An Hour. The Architecture Is Starting To Pay Off

A useful architecture test Qwen3.8-27B became executable in C-Kernel-Engine (CKE) in roughly an hour of focused bring-up work. The important part is not the clock by itself. CKE did not need a new arithmetic kernel or a Qwen3.8 branch inside the compiler. It needed an explicit model identity, a dense circuit, and the kernel contracts we had already hardened while working on Qwen3.5 and Qwen...
CKE Zip: From Splitting Tokens To Splitting Qwen3.5 Weights

CKE Zip: From Splitting Tokens To Splitting Qwen3.5 Weights

The axis matters CKE's first two-node Qwen3.5 experiment divided prefill token rows while both machines retained complete weights. PR #382 now tests the axis we actually wanted: attention and MoE weights are split across the Ryzen and Intel P3, each node computes a partial result, and the full transformer block is reconstructed through explicit reductions. Yesterday I wrote about the smal...
I Spent CAD $9,575 Building CKE's CPU AI Lab

I Spent CAD $9,575 Building CKE's CPU AI Lab

A small CPU AI lab is now real I have personally spent CAD $9,575.16 on the computers, memory, and storage now supporting C-Kernel-Engine. The important part is not that I bought hardware. It is that CKE can now run the same model circuits across an Intel AVX2/VNNI server, a Zen 5 AVX-512/BF16 server, a portable development machine, and an old compatibility node. The next job is to turn tha...

Training vs Inference in CKE: One Interface Discipline For Both Lanes

Today I want to write about how CKE does the thing I care about most: it gets better at saying no. Not no to models, and not no to speed. No to bad wiring. No to quiet mismatches that used to slip through and only show up later when the result was already wrong. If you do not know CKE, start with What Is the C-Kernel-Engine and then the earlier notes on templates as circuit maps....

The Qwen3-VL CPU Gap vs llama.cpp: A Build Log That Keeps Getting Measured

I am revisiting this gap today because it is the measurement I keep being asked about, and because the honest answer has changed shape since I first wrote it down. The numbers below 281 seconds down to 76.7 seconds on the Qwen3-VL OCR path, against a llama.cpp baseline around 55.8 are the historical record from early July, and I am leaving them untouched. What machine produced the

How CKE Picks a Kernel Without a Single if-Statement: v8 Provider Selection

Inside the CKE compiler One CKE runtime serves many model families, quantization formats, CPU instruction sets, and execution phases. So how does it decide which exact C kernel runs for each operation? The answer is 285 machine-checked JSON maps and a resolver that fails closed never a Python if model == "qwen3vl". This post walks through that decision, with real eviden
Antsand vs WordPress, Webflow, Ghost, And Substack: Five Different Platform Contracts

Antsand vs WordPress, Webflow, Ghost, And Substack: Five Different Platform Contracts

Yesterday I wrote about what Antsand is, why I kept building it for fourteen years, and where I want to take it. The obvious follow-up question is simpler: why did I build any of this when WordPress, Webflow, Ghost, Substack, and a growing number of AI website builders already exist? The honest answer is not that Antsand is better than all of them. In several areas it is clearly wo...

What Is Antsand? Why It Took 14 Years and Where It Is Going

Research note Antsand is the platform that runs ShivasNotes, my client sites, and my publishing workflow. This post explains what it actually is, why it took fourteen years, and where it is going including how it is designed to become a data surface, an agentic surface, and eventually a training and compute surface as C-Kernel-Engine matures. Today I want to wri

Linux Memory Allocation: malloc, mmap, Pages, TLBs, and Huge Pages

I keep allocating very large buffers in CKE, but malloc returning a pointer does not mean those bytes are already sitting in DRAM. It means the allocator has promised me a range of virtual addresses. The physical pages the actual DRAM show up later, and when they show up is a choice I get to make as a runtime author, whether I make it deliberately or by accident. For a long time I

Training vs Inference in CKE: One Interface Discipline For Both Lanes

Today I want to write about how CKE does the thing I care about most: it gets better at saying no. Not no to models, and not no to speed. No to bad wiring. No to quiet mismatches that used to slip through and only show up later when the result was already wrong. If you do not know CKE, start with What Is the C-Kernel-Engine and then the earlier notes on templates as circuit maps....

A Green Checkmark Should Name Its Machine: How CKE's Nightly Report Became Evidence

I opened our nightly test report this morning and caught myself doing the thing I always do: glancing at the green badge, nodding, moving on. And then it hit me that the badge had been telling me almost nothing. One hundred and fourteen checks were green on whose machine, with what instruction set, proving what, exactly? I spent the day making the report answer those questions itsel
LinuxUtilities: My Personal Linux Control Layer For AI, Publishing, And C-Kernel-Engine Work

LinuxUtilities: My Personal Linux Control Layer For AI, Publishing, And C-Kernel-Engine Work

Today I want to talk about a collection of small Linux utilities I have been building to support my daily work. Since last year, and through most of this year, Linux has become my primary operating system. Almost all of my AI, C-Kernel-Engine, Antsand, writing, and presentation work now happens there. I have used Linux for many years, but usually through servers, VirtualBox, or ano...
How CKE Made Whisper Faster In Two Days — And The Evidence That Makes It Believable

How CKE Made Whisper Faster In Two Days — And The Evidence That Makes It Believable

The two days after the July books closed The July recap closed its Whisper chapter at a 2.7x gap to whisper.cpp: 1.531.58 seconds on the controlled Base fixture against 0.567 seconds for the reference. In the first two days of August, three more constraints fell FP16 attention and scheduling, FP16 GEMM weight reuse, and the stride-two Conv1D gathers. The first rung mov
What CKE Built In July 2026: Vision, Audio, X-Ray And Faster CPU Inference

What CKE Built In July 2026: Vision, Audio, X-Ray And Faster CPU Inference

July 2026 engineering recap CKE merged 194 pull requests and 466 commits during July. The useful story is not the volume. It is how the work connected: Qwen3-VL numerical debugging forced us to make circuits, kernel maps, runtime provenance and X-Ray more explicit. Those changes then made it easier to add Whisper audio, harden hybrid language models, and optimize Qwen3.6 with...
How CKE Made Qwen3.6 Prefill 2.6x Faster

How CKE Made Qwen3.6 Prefill 2.6x Faster

The performance work after numerical debugging CKE first used X-Ray to find why Qwen3.6 and llama.cpp diverged. Once the recurrent topology and arithmetic contracts were bounded, the same evidence showed where prompt processing was slow. The next question was harder than simply adding threads: where can we run work concurrently without changing the model's numerical trajector...
Modelling A Quadcopter: From Four Rotor Speeds To Motion

Modelling A Quadcopter: From Four Rotor Speeds To Motion

Revisiting an older quadcopter video I made the video below while trying to understand how four rotor speeds become lift, yaw, pitch, roll, and eventually motion through space. This written version gives me room to slow down the explanation, redraw the important relationships, and correct a notation mistake that viewers caught in the original video. .quadcopter-video-...

Archived Personal Posts

Older writing, daily notes, business posts, year-end notes, and personal archives.

The 3rd Week of January

personal 3 min read
17 Jan 2021 by Anthony ShivakumarThis past week or so I have been doing good energy-wise. One being, I can get up very early. Anywhere between 4:00 am and 5:00 am. I do not put an alarm. I get up organically. I love it. After I get up, go to the washroom and wash up, I then sit and do the Wim Hof breathing exercise. It is a great way to start the day. A bit of deep breathing, yoga, and some medita...

Client work and system identification

personal 1 min read
On Monday I was working on a bit of client work and on drone tech. I woke up at 5:00am and started to do week 2 of the wim hof breathing. It is a good start to my morning. Get my lungs moving and helps set my mind for the day ahead.As mentioned, mostly client work was the goal of the day. Monday and Tuesdays will be focused more on client work and ANTSAND, while the rest of the week will be focuse...

Understaning the Kalman Filter

personal 3 min read
I am writing my Saturday and Sunday Journal together and a day later.Saturday the 9thOn Saturday the 9th, I got up arund 6:30. I haev not done the Wim hof Breathing ever since I did my afternoon run and got a bad eye and head ache. I just wanted to take a day or two off from breathing. Which in the morning allowed me to spend time working. I had to reply to a friend which I did. It took about a
Headache and eye hurt

Headache and eye hurt

personal 2 min read
Today started off well. Got up at 4:45 and started to do my Wim Hof breathing. Then I got to work on the sensor fushion algorithm. I had breakfast at 7:45am. Around 8:30 am I had a shower. I ended it with 1min of cold water. It felt good. I still believe I have the groin cycst. The cold shower helped a bit.After which I had a meeting at 9:30 am and again go to work on the sensor algorithm. My g

The importance of priority #1. What is ANTSAND and why is it being built?

Business personal 7 min read
ANTSAND first started as my personal dairy to keep track of my life. At one point in time, I did not wanted to use any Social media sites. I wanted to only use the one I created. I think the creator in me thought, if I can build one, I will build one. Why do I need to use other products. This idea has still not completely grown out of me. For the most part I am vey much a maker.I have been working...
Progress on Drones

Progress on Drones

personal 2 min read
I did not write yesterday's journal. Today and yesterday I woke up at 430 am. This is so that I can squeeze in the Wim Hof Method in the morning. I want to get into the practice of breathing, meditating yoga, and perhaps even cold showers. My eating habits too, have been improving. Though today I did skip lunch. I had a slight headache and my left eyes were hurting. Plus I woke up at 4:30 am so I ...

3rd day of the year

personal 2 min read
Today, I got up at 5:00. Started to do the wim hof method week 1. It is going good. I feel great after each session. I want to get up earlier, perhaps at 4:30a.m so that I can get a head start on the meditation, breathing and yoga.At 7:30 a.m I cooked Uppma for breakfast. Eating breakfast in the morning is a wonderful routine. It help me keep check of the time. This way I am aware what I am doi

The 2nd day of the year

personal 1 min read
Today i woke up at 5:00 am. around 5:30 am or so I started to do the Wim Hof method. The breathing exercise, yoga and meditation. At 7 am I had to go out to get some bread for my breakfast. I then came home and prepared breakfast.For breakfast I prepared a sandwich, avocado, roasted chic peas and sprouts. With the sandwich I also had leftover butternut squash soup. It was a good healthy breakfast....

1st Jan 2021

personal 3 min read
I am not big into the New Year celebration thing. It is just another day. Every day is kinda a new year. So, it is just a matter of perspective. Nevertheless, the date change does make a psychological impact. the 1st of the month, a different or new year. It is in these times we tend to set goals, resolutions, change things up for the better, make commitments, and try to keep up. I am no differ

A cooking day

personal 1 min read
Woke up around 6:00 am. and worked on making a proper meal plan. We then went for gorcery shopping at 8:00 am. After we came back I made myself brunch while washing some of the fruits we just bought. I did not get much work done today. I did take a short afternoon nap. I may need to drink coffeee to keep my self awake. After I woke up, I prepared to cook the butternut squash. While cooking I did m...
Working on ANTSAND

Working on ANTSAND

personal 2 min read
Today is ANTSAND day. Monday and Tuesdays is where I will dedicate my time on ANTSAND. I got up today around at 6:00am. Went to the toilet, brushed my teeth and got to work. Around 9:45 am or so, I did my TACFIT moderate intensity workout. It is always good to follow and stick to my fitness plan. In the coming new years, I will want to make fitness and my eating habits set. No more burping, fartin...

A lazy day

personal 1 min read
Today, I got up early and worked on some client work. I felt very bloated in the morning. A lot of burping. I also sent a mail to a potential co-founder on starthawk. I most likely will need to look for a co-founder for ANTSAND. Will need someone who can take care of sales and marketing.We then went for a walk followed by a run round the park. The distance covered was approx. 2.7 km. I beleive
Root Vegetables

Root Vegetables

personal 2 min read
Today, the 27th of December (Sunday) was a normal day. Got up around 7. We did our recycling before I joined my mastermind. The mastermind session was good. Chill, relaxed and we just spoke on few different topics. One of the books that was recommended was Siddhartha y Herman Hesse. Another interesting blog that was mentioned was neilonsoftware.com .After mastermind, I did feel a bit lazy an

Getting the acclerometer working

personal 1 min read
Today, once again I got up early and started to work on getting the accelerometer working. I did start to get it working. Had a meeting with a friend that is likely to collaborate with me to make drones. After the meeting, we went for a walk. I had my mastermind meeting today too. I only spent an hour in the meeting. don't want to spend more than 3 hours a day on meetings.After which I did m
I2C and learn how to troubleshoot

I2C and learn how to troubleshoot

personal 1 min read
Today is hardware day.. Woke up early and got working on interfacing the gyroscope and accelerometer. The whole morning went trying to get I2C working. The signal were just not right. It took me the whole day to realize that there was a connection isssue. Nevertheless, I got it working. A clean signal like this always feels good.Healthwise, I am feeling a lot better. No more pain as such. I did do...

Learning Kalman Filtering ...

personal 1 min read
Today was a lazy. Woke up a bit late, did not do any exercise and just not productive as such. Towards the evening I did spend some time learning how to use Kalman Filtering for attitude control of a quadcopter. Tomorrow, I will try to get the accelerometer and gyroscope working and see if I can start to apply the Kalman Filter.My stomach still has the left side discomfort. I did go to the toilet ...

Recovering ... but bloated

personal 1 min read
Today, I did not want to do too much. So I spent time studying some Machine Learning Algorithims. Pretty much the whole went in learning.My health is better, though I felt really bloated after dinner. I ate a decent size meal, but the bloating is a bit concerning. Hopefully the tablets I take Ilaxatives) will help relieve some of these issues.Tomorrow, I might spend more time on ML. Perhaps wor

Recovering ...

personal 1 min read
Today, I was feeling much better. I went to the doctor and they gave me a few tablets. If the problem still continues after 2-3 weeks then I need to go back. I have got most of my appetite back. I was able to eat my salmon, kale, yoghurt, naan, sprouts and avocado for lunch.I plan to now work, 3 days on hardware and 4 days on software. The 4 days of software will include working on A.I, machine

Little Tension and a christmas tree

personal 2 min read
Today, I did not sleep too well. I work up at around 3:30 a.m. due to burping and farting. I did drink the prune juice yesterday. It might have been the cause of my discomfort. My fiance could not bear the smell, so she slept in the living room. After an hour or so, I did got to the toilet and was able to pass motion. The prune juice helped. When I got back to bed, I did not want to start worki

let the progress begin ...

personal 2 min read
Alright, today was just a normal day. I work up at 6:30 am. I did pass motion, so that was good. Still constipated though. Not sure what the issue is. Hope it is not something too critical.We went for a morning walk as it was bright and sunny. A walk always helps. I might need to walk more often to get my bowel movements going. Either in the morning on after lunch.Today, was also the last day o

Another slow day

personal 1 min read
Embedded system programming is tedious. Sometimes things work fast and quickly, sometimes they just take time. My progress this month has been very very slow. I have switched from working on the nrf24ap2e to the nrf24l01+. The latter is easier to configure. I want to get the radio working by the end of the month. .. Hopefully I can get something working soon.My fitness is going well. I have com

The day of ups and downs

personal 2 min read
Oh, today was an exhausting day. Not physically, but emotionally.My day started like any other day. Woke up early, passed motion, which is good. Not too constipated. Had a shower and got to work. Today is solder day. Which is, the day to solder another radio and get the two radios talking. Unfortunately, the soldering did not go as well as I planned. In 2014, I had bought enough components to make...

Let us get moving

personal 1 min read
Today, once again i started the day early. A bit constipated. Hopefully nothing major. I might have to do more movement exercises. If it persists for a few more days or a week, I will have to go see a doctor.Other than that, I got most of the ANT+ radio config working. I was not able to check over the air signals. I will need to solder a new radio tomorrow or by the end of the week. I also was abl...

More work .. but progressing

personal 1 min read
Today, once again I got up early, around 5:30 a.m. Had a shower and got to work. My goal for today was to progress more on the radio. I want to get it to start interfacing. Te protocol is a bit complicated. I hope to get it working by the end of the month.I also studied a bit from the course "Sensor Fusion and Non-linear Filtering for Automotive Systems" on Edx. It seems to be a decent course. I w...

Drone Specification ...

personal 2 min read
Today, was just another day. Got up a bit late, around 6:30 a.m. I did a bit of the machine learning course in the morning. After which, I did my workout. Before my workout, I was once again browsing the different courses on Coursera / Edx related to image processing, sensor fusioh etc. It is easy to get carried forward in designing a very big system. To stop my self from doing that, I spent ti

Getting the radio to work ...

personal 1 min read
Today, once again I got up early. Not too early, around 6:30 a.m. My plan was to work more on the radio and get it interfacing with the microcontroller. I am progressing well in this regard. Hopefully soon, I will be able to send data over the air. This will be a good achievement.I also had our mastermind session today, The discussion were good. We talked about business, ideas, philosophy and r

Another Day ...

personal 2 min read
Today, once again I got up a bit early, at 6:30 a.m. Had a shower and worked a bit on ANTSAND. There are bugs in the app, and I plan to work on it frequently. One thing I have realized today, regarding the making of the drone, is to not focus on theoretical stuff anymore. The only theory I can work on is on A.I. I will only get back to theory once I have a drone that is flying and working. It is i...
Let's get the drones sketched

Let's get the drones sketched

personal 3 min read
Today, once again, I got up early, around 5:30 am, had a shower, and got down to work. In the morning, I was working on interfacing the radio with the micro-controller. I got most of the device driver working. Next, I will need to send some ANT+ commands and see if it will transmit the messages over the air. After which, I did my Tacfit Commando. Today, was a high-intensity day. I did push myself ...
A dumb A.I.

A dumb A.I.

personal 3 min read
Today, my day did not start at 5:30 am. It started at 2:30 am when a funny smell polluted our room. We woke up from the smell as it smelt like burnt plastic. We were not sure if it was caused by something in-house or it was coming from outside. We went out to check and the smell was stronger outside. There was nothing we could do, so we closed all the windows and went back to sleep.Sleep interrupt...
Getting Closer

Getting Closer

personal 1 min read
Once again, today I got up early, had a shower and got straight to work. The agenda for today was to work on the radio fro 6:00am to 10:00 am. I did stick to my plan. I was able to progress further with the radio. Hopefylly by the weekend, I can get it to work properly.After working on the Radio, at 10:30 or so I did my Tacfit Commando. The exercies, the morning routine and the evening routines

Getting the radio working - Work in progress

personal 3 min read
Today, once again I got up early at 5:30am, had a shower and got straight to work on interfacing with the radio. Embedded systems and interfacing is not particularily difficult, but it does cosume a lot of time. Espcially if you are trying to get something working with limited resources. When it comes to these types of things, the manufactuere should provide sample code. In this case, little to no...
Let the training Begin

Let the training Begin

personal 4 min read
My morning routine of getting up early is going well. I got up at 5:30 am, had a shower, and got straight to solving the backpropagation algorithm. Surely, in the mornings my mind reaches a new level of concentration and focus. I got the assignment working in an hour. After which I did my Tacfit commando. Today was the high-intensity day. I believe I pushed myself quite hard. I reached a heart rat...
Back Propogation

Back Propogation

personal 1 min read
Today, once again, I woke up at 5:30 am, had a shower, and got straight to work. Today I wanted to get the back-propagation algorithm of a neural network to work. Unfortunately, I am still struggling to get it to work. What I did today?I woke up earlyI did my Tacfit Commando exercisseI went for a walkI took a good afternoon nap.I worked on Week 5 of the Machine learning course by Andrew Ng. I am ...
Neural Networks and a so so day ...

Neural Networks and a so so day ...

personal 3 min read
Today, once again I got up early, around 5:30 am, had a shower, and got down to work on completing the neural network chapter from the course Machine Learning on Coursera by Andrew Ng. I am now in chapter 5 and need to finish the backpropagation chapter. Once I get the backpropagation to work, I will start to train the neural network to create websites. At least, this will be the 1st step towards ...
The Radio

The Radio

personal 5 min read
Today, once again, I got up early, around 5:30 a.m. I had a shower and got straight to work. My goal for today was to get as much client work done by 10:00 am, do a workout, and start working on interfacing the radio module. I had a bit of pending client work, so I tried to get as much done. At 10:00 a.m, it was a hard stop. This meant, even though I did not complete the client work, I had to stop...
Client work

Client work

personal 2 min read
Today was client day. I got up early at 5:00am, had a shower and got to work on client work. I am working on a website for a clinic. The design needed a few elements that were not readily available in ANTSAND. so I had to progam a few features in. I worked for a few hours before I had a call with my business partner. After which I did my cycle 3 of tacfit commando. The fitness side of things is go...
Machine Learning progress

Machine Learning progress

personal 3 min read
Today seemed to be a good day. I got up early, around 5:30 am, brushed my teeth, had a shower, and got started working on the machine learning course by Andrew Ng. I wanted to complete the logistic regression chapter and get started with a neural network. I wanted to end the day implementing my neural network to auto-create websites. Unfortunately, I procrastinated a bit. Plus, it does take time t...
The Starry night

The Starry night

personal 3 min read
Today, I got up a bit late. Around 6:30a.m. The reason being I slept a bit too early last night and had to get up in the middle of the night. The sleep interruption offset my sleeping schedule by an hour. After getting up, had a shower and got back to work. I posted my app on product hunt and submitted ANTSAND to app sumo. A bit of marketing. Then, I did my tacit commando. After which I needed to ...
logistic regression

logistic regression

personal 1 min read
Today was a slow day. I did get up early around 5:30, had a shower and got to work. I completed a few client work in the morning. I also did my Tacfit commando workout. After which I studied logistic regression from a machine learning course taught by Andrew Ng from Coursera. Around noon I cooked some fish curry and ate the same for dinner. Noon time I was once again studying the course. Tomorr
The pivot

The pivot

personal 2 min read
Today was a good day. I got up early, had a shower, worked a bit on ANTSAND. Did my Tacfit Commando (high-intensity day). Then we went for a walk around the nearby park. The walk was half-hour long. After this, I had my mastermind meeting from 10:00 am to 12:00 pm. I needed to install OpenCV to do some image processing/ computer vision work. I had a few issues trying to install it. Once I got it t...
Headaches and a dud day

Headaches and a dud day

personal 2 min read
Today was a slow start. Got up around 7 a.m. We went grocery shopping around 8:30 a.m. As the covid cases are rising, we were preparing to stock up on some food in case there will be another lockdown. After coming back from grocery shopping, I had a slight headache. I took a nap before having a call with one of my mastermind friends. We talked about drones and more about the ANTSAND app and how he...
the Dilemma

the Dilemma

personal 2 min read
Today, once again I started the day early. Had a shower and tried to figure out what I want to do today. I wanted to get started doing some machine learning A.I. course. I also needed to work on getting the infrastructure in place. Which is the ability to get data, process data, train the A.I., and then deploy the A.I. to an application. There is a bit of backend code that is needed to be done. Th...
Just another day

Just another day

personal 2 min read
Today, once again I started the day by getting up early, around 5:00 a.m. Had a shower and got down to work on client work. I needed to add some animation on a client website. Since all my clients website are now directly powered by ANTSAND, I had to fix a feature in ANTSAND itself. I got it working so that is good.Then around8:30 am I did my conditioning exercise. Day 5 of my program. Today was a...
A sore body and a semi-productive day

A sore body and a semi-productive day

personal 3 min read
My morning routine is going great. I got up at 5:30 a.m, had a shower, and started to get down to work on antsand. At 7:00 am, I had a consulting call with one of my clients. At 8:15, the call ended and I got down to do my high-intensity workout. The 1st cycle of mission one is complete. Six more to go. The table below shows my performance.Day -4 Mission 1 Tacfit Commando - High IntensityExercise1...
Let the exercise begin

Let the exercise begin

personal 4 min read
Today, I started the day pretty well. I got up at 5 a.m., had a shower, and got right down to work on ANTSAND. Today's goal was to fix SSL or auto-configure websites for security using LetsEncrypt. Honestly, all my design ideas turned out to be fruitless. Which is to say, I would need to come back to this at a later date and figure out how to auto-configure websites for security. One of the reason...
Day 2 of my Journaling

Day 2 of my Journaling

personal 2 min read
Today is my 2nd day of writing. I got up early, around 5:00 am. Daylight savings did help me to get up early. Had a shower and was ready to get started programming. Being the 1st working day of the month, I sent invoices to my clients. Today, I had to work on the form section of ANTSAND. Finish up with the "don't break the chain rule" feature. While doing so, I got a chance to refactor some of my ...
Blog writing and habit tracking

Blog writing and habit tracking

personal 3 min read
Today I have started a new blog called shivasnotes.com. This blog is to help me keep track of my journaling, habits, and accountability. I have wanted to do this for a while, but I kept coming up with excuses. Nevertheless, I am happy I got started.In this blog, I plan to write on my challenges like the 30-day challenge, rant on random topics like productivity, health, fitness, hobbies, mindsets, ...
39+ mental tools I use to make better decisions

39+ mental tools I use to make better decisions

Are you looking to start a company, make a drastic change in your life or are you stepping into the unknown? There are forces of nature that are there to help and guide you if you look for them. It can get scary, fearful, and very daunting. As a result, I have compiled a list of things that will guide me even in the darkest moments of life and business.   This is a list of belief that I use as a
13 things I learnt after doing the 10-week wim-hof method

13 things I learnt after doing the 10-week wim-hof method

lifestyle personal 14 min read
Do you get a migraine or sinus and lose a full days worth of productivity? You think yoga is for old people or it is too boring to do it? You tried meditation, and think you are better off going to sleep than sitting like a fool? Or even the thought of cold showers frightens you, why would anyone do it?  Let me tell you a little about myself. I have lived with headaches like a migraine an

Subscribe

Get my rants delivered to your inbox

I will send new posts as and when I write. No fixed cadence, just engineering notes, rants, and things I am thinking through.