AkitaOnRails Blog
Featured
LLM Benchmark v4: the open source verdict (as of Oct 2026)
Almost 70 runs put the best open models close to Sonnet 5. Even so, 159 tok/s on an RTX 5090 is still slow for my work, while faster options already cost as much as a server.
We're Living in the Future: Cars Fly and We Don't Have to Write HTML/CSS Anymore
Tesla's Roadster spun in midair while I redesigned six sites in about a day. With agents handling web, mobile, and desktop, mechanical work became automation; careers demand adaptation.
I Trained My Own AI Model (Spoiler: It's Just for Study)
One exercise was missing: training an LLM from scratch. I started from Karpathy's microgpt, 200 lines of pure Python, and expanded it into a lab with a tokenizer, pretraining, SFT, an OpenAI-compatible server and OpenCode integration. Under two weeks, one RTX 5090, 10 billion tokens, 137 hours of training, at most around R$ 80 of electricity. The final model has 337 million parameters, gets 58 out of 100 simple Python functions right and scores zero on anything that requires acting in a repository. I explain each step, the stumbles, the dimensions against GPT-2 and frontier models, and why I think the future is smaller, specialized models.
I Self-Published My Own Book/e-book: Competent Enough
I published my book about AI, for free, in EPUB and PDF: the blog's 131 posts from 2026, almost half a million words, reorganized into 20 chapters, 10 appendices, a glossary and more than 300 references. This text is the technical backstage: the inventory with a hash per post, the contract every agent read before writing, the parallel drafts, the invisible per-paragraph marker that ties the book to the blog, the checker that fails a chapter with an error, what the fact-check caught wrong in my own posts, and the Quarto, Typst and Pandoc pipeline that generates everything. It's a preview: open an issue if you find a mistake.
AI Reverse Engineering: Is Adobe Dead?
In the same week, a dev dropped seven Adobe clones built with Claude, REA passed 100K stars promising to 'reverse engineer anything', and retro gaming got a new port every day. The viral conclusion: no binary is sacred, Adobe is dead, programmers are dead. I explain what each of these things actually does: PhotoCraft is an alpha with 800 open issues, REA is a toolbox that requires the binary and Ghidra, and retro moves fast because an 8 MB ROM turns into 135K lines of C. Rewriting Bun cost US$ 165K with the author and 64 agents; Photoshop doesn't even have an open binary in the parts that matter. Reverse engineering got easier. Magic still doesn't exist.
LLMs + Agents = competent enough
A synthesis of what I've been measuring and writing: LLMs converged. The transformer architecture got near the ceiling, the parameter war is over, and the leap everyone felt in 2025 and 2026 came from the agentic harness (tool calling, prompt caching, long sessions), not from a base intelligence orders of magnitude larger. I retell the timeline from 2019 onward with sources, explain why the agent allows a less dense model, and why that's already competent enough to automate real software engineering.
The Limits of LLMs: Good at Reproducing What Exists. Expensive for What Doesn't
Every LLM was trained only on what's public, but 81.5% of GitHub activity happens in private repositories, and the source code of every proprietary software never made it into training. In my retrocomputing projects I felt firsthand where the machine stalls: with no reference and no oracle, it spins in directed brute force.
Watch Out! A Pseudo-TechCrunch Tried to Scam Me! And Calendly Is SHIT
An account posing as a TechCrunch journalist DMed me today asking for a meeting, with a Calendly link that redirects to an X OAuth authorization screen. It's the same scam that already caught José Valim, Elixir's creator. I document the scam step by step, with technical proof.
Stop Making Excuses and Ship More! With AI, the Premise Changed. Get It.
With frontier LLMs making fewer mistakes than most human devs, the only real excuse left for holding back a deploy is your own review process's bureaucracy. I use my open source projects, cross-platform CI/CD, and real contribution numbers to show how I've been doing this since February.
AI Unmasked the Climate Change Hysteria
AI's race for energy buried the climate agenda: big techs restarting nuclear plants, Germany admitting the mistake, datacenters already consuming double the global EV fleet. With Bjorn Lomborg's numbers, the climate apocalypse as a misallocation of trillions.
Talking a Bit About My AI Skills
Agent skills are just prompts in text files. I keep 25 in my my-skills repo, from pr-audit to fact-check, and they have already closed over a thousand PRs and issues across my projects. The recommendation: don't copy somebody else's skill, build your own.
You're an Idiot if You Believe the Misleading Ads from OpenAI, Anthropic, NVIDIA. Here's Why
Nvidia and OpenAI declared that AGI has arrived, OpenAI brute-forced a Millennium Problem and threatened a mathematician's career, and an Anthropic employee quit crying Skynet. I take the propaganda apart.
AI Challenge: Converting NES ROMs to the Master System/SMS
I tried using frontier LLMs to convert NES ROMs into Master System games. I explain the architecture, why translating 6502 to Z80 runs slow, the emulator as a feedback oracle, and where the mappers stall everything.
AI-MEMORY 2.0 - The Best Memory System for Agents and Teams
ai-memory hit version 2.0 with the open OKF format, local embeddings on by default, and support for teams working in parallel. I compare it with the competition and show what only ai-memory delivers.
Hot Take: Harness, Loop Engineering, Graph Engineering Are Bullshit
I posted that Harness Engineering, Loop Engineering, and Graph Engineering are bullshit to sell consulting hours and courses. Here's my receipt: 30+ public repos, an AI marathon, and a whole benchmark — none of it using any of those 'disciplines'.
Quantum News: Majorana 2 and Understanding Shor
Microsoft announced Majorana 2 with 20-second qubits and physicists answering that nothing was resolved. I take the chance to properly explain Shor's algorithm, why factoring becomes period finding, and what quantum computers are actually good at.
Using AI to Solve My Little Day-to-Day Problems
A roundup of my small open source projects: a desk clock with widgets, manga readers, decent email, typing practice, YouTube karaoke, ComfyUI in Docker and more. All born from real little problems in my day-to-day.
How Do I Protect Myself From My Agents Deleting My Stuff?
LLMs deleting files from famous people made headlines this week. In five months of heavy use, in YOLO mode, it never happened to me. But I don't trust them either: BTRFS snapshots, restic backups, sandboxing and discipline.
Why LLMs Will Fail at Your Company
The deliberately speculative thesis is that LLMs amplify processes where people outsource decisions. The proposed alternative is daily cycles of decision, implementation, testing, and review close to the outcome.
AI Controversy in Open Source Project Contributions - My Take
AI in open source contributions is here to stay, even with AI slop, regressions, and false bugs. The practical answer is to automate triage and auditing without taking the final decision away from humans.
Open Source Best Practices with LLMs - The Bare Minimum
Open source projects built with LLMs are only ready with simple installation, reliable tests and CI, and documentation focused on the problem. Standardizing releases and deployments makes automation predictable.
Clean Code for AI Agents
Clean Code carries different weight when an agent is the primary reader: small code, greppable names, provenance context, headless tests, and explicit rules reduce navigation, cost, and errors.
VS Code Is the New Punch Card
In the era of coding agents, typing everything by hand in VS Code is becoming the modern equivalent of a punch card. Foundations did not become legacy.
RANT: Did Akita Bend Over for AI??
I have been 'bending over' for AI since 2023. If you only watch out-of-context podcast clips, let me spell it out for you.
Vibe Code: From Zero to Production in 6 DAYS | The M.Akita Chronicles
How I shipped a full newsletter, blog, podcast, and Discord bot to production in six days using vibe coding the right way.
RANT: Did AI Kill Programmers?
Akita's rant on why AI won't replace real programmers, why the programming bubble already popped, the S-curve ceiling of current LLM architecture, and what actually changes for engineers in the LLM era.
AGI or Skynet Isn't Coming Anytime Soon
Why the current LLM architecture cannot reach AGI, and why the hype from Big Tech CEOs and media-anointed experts is mostly backroom negotiation noise.
RANT - LLMs are LOOT BOXES!
Why LLMs for real coding behave like gacha loot boxes, and why every incentive in the industry pushes you to burn more tokens.
Recent articles
October 2026
17 articlesWhy Don't We Write Definitive Formal Proofs to Get Bug-Free Code?
Anthropic proved 123 open-source libraries never crash, using 180 million lines of AI-written proof. What formal verification is, why seL4 cost 20 person-years, why Bend 2 changes the game, and what PHP2Shell, GitLab and Railway say about technical debt coming due.
LLM Benchmark v4: the open source verdict (as of Oct 2026)
Almost 70 runs put the best open models close to Sonnet 5. Even so, 159 tok/s on an RTX 5090 is still slow for my work, while faster options already cost as much as a server.
Best Open Source LLM: Strata with Qwen 3.8 Flash Next
Qwen3.8 Flash-Next served by the Strata engine scored 91, Tier A, on an RTX 5090, 1.5 points behind the B300 and at about 159 tokens per second. I explain how Strata fits a 125B MoE on a PC by using VRAM as an expert cache and RAM as the warehouse, why that matters with memory costing twice as much, and why it isn't magic: the 5090's memory bandwidth is seven times the Strix Halo's and almost 30 times PCIe's, and even on my machine it takes up almost all the RAM and VRAM. I show the install, a service that loads and unloads the model on its own, the OpenCode configuration, why I consider my machine the minimum for real programming work, and why, even a bit less responsive than a commercial model, it's already almost good enough to replace the cloud day to day.
LLM Benchmark v4: consumer-grade open source on RTX 5090 or Strix Halo. And Strata?
I went back to consumer hardware after the datacenter tests. Qwen3.8 Flash-Next scored 91.0, Tier A, on an RTX 5090 with the Strata engine, 1.5 points behind the B300 at 159 tokens per second. It's a different model from the Qwen 3.8 27B I had called slow: a MoE with 6B active per token, served with the hottest experts on the card. Strix Halo scored 80.0 with the 27B, but at 7.7 tokens per second. It's the first local model to make Tier A on the v4, and it's worth testing in day-to-day programming, if you can find a 5090: NVIDIA denies the end of production but admits tight supply.
LLM Benchmark v4: can my team fit in a 4x b300/h200?
After ranking the big models, the leftover question: how many developers fit on a 4x B300 or 4x H200 pod? We measured real concurrency, with a latency SLO, and the answer changes everything: DeepSeek V4.1 Flash holds 32 comfortable users on B300 and only 8 on H200. Cost per comfortable user: US$ 1.13/hour on B300 versus US$ 1.80 on H200. When your own infra closes the math against a frontier subscription, and when it's just expensive vanity.
LLM Benchmark v4: open source that only fits in 4x B300!!
I rented 4x NVIDIA B300 on Runpod and ran my v4 sabotage benchmark on the open models that don't fit on a single card. DeepSeek V4.1 Flash, at 748B, tied the leaders at 92.5, on the level of hosted GLM 5.3 and GPT-6 Sol. GLM 5.3 full fell behind, the 1T Kimi K2 Thinking disappointed, and a 397B model plummeted from 94 in the cloud to 27 at home. The math: US$ 36 per pod hour, US$ 864 per day. When owning your own infra makes sense, and whether a big open model is already enough for your day-to-day programming.
LLM Benchmark v4: the top open source LLMs on a B300
I rented a 288 GB NVIDIA B300 on Runpod and ran my v4 sabotage benchmark on 14 open-weight models that don't fit at home. None of them serves fast the way it ships: GLM 5.3 Flash went from 10 to 141 tokens per second on configuration alone. Two open models reached 92.5, Sonnet 5 level, and a 27B fine-tune scored 88. A 397B flagship let an admin backdoor through, DeepSeek V4 Flash scored 86 in the cloud and 44.5 at home, and Nemotron cheated on a test. The math: the cloud API of the same model costs a dollar per run and the pod costs 218 a day. When owning your own infra is worth it, and whether a team fits on one card.
We're Living in the Future: Cars Fly and We Don't Have to Write HTML/CSS Anymore
Tesla's Roadster spun in midair while I redesigned six sites in about a day. With agents handling web, mobile, and desktop, mechanical work became automation; careers demand adaptation.
I Trained My Own AI Model (Spoiler: It's Just for Study)
One exercise was missing: training an LLM from scratch. I started from Karpathy's microgpt, 200 lines of pure Python, and expanded it into a lab with a tokenizer, pretraining, SFT, an OpenAI-compatible server and OpenCode integration. Under two weeks, one RTX 5090, 10 billion tokens, 137 hours of training, at most around R$ 80 of electricity. The final model has 337 million parameters, gets 58 out of 100 simple Python functions right and scores zero on anything that requires acting in a repository. I explain each step, the stumbles, the dimensions against GPT-2 and frontier models, and why I think the future is smaller, specialized models.
I Self-Published My Own Book/e-book: Competent Enough
I published my book about AI, for free, in EPUB and PDF: the blog's 131 posts from 2026, almost half a million words, reorganized into 20 chapters, 10 appendices, a glossary and more than 300 references. This text is the technical backstage: the inventory with a hash per post, the contract every agent read before writing, the parallel drafts, the invisible per-paragraph marker that ties the book to the blog, the checker that fails a chapter with an error, what the fact-check caught wrong in my own posts, and the Quarto, Typst and Pandoc pipeline that generates everything. It's a preview: open an issue if you find a mistake.
AI Reverse Engineering: Is Adobe Dead?
In the same week, a dev dropped seven Adobe clones built with Claude, REA passed 100K stars promising to 'reverse engineer anything', and retro gaming got a new port every day. The viral conclusion: no binary is sacred, Adobe is dead, programmers are dead. I explain what each of these things actually does: PhotoCraft is an alpha with 800 open issues, REA is a toolbox that requires the binary and Ghidra, and retro moves fast because an 8 MB ROM turns into 135K lines of C. Rewriting Bun cost US$ 165K with the author and 64 agents; Photoshop doesn't even have an open binary in the parts that matter. Reverse engineering got easier. Magic still doesn't exist.
Linux users: always choose BTRFS, not EXT4
Sequel to the NAS disaster article: filesystem choice is a life-or-death decision for your data. The history of ext4 and btrfs, snapshots, subvolumes, reflink deduplication, corruption protection, compression measured in real GB (630 GB free on 2.2 TB), ZFS vs btrfs on NAS, snapshot boot recovery with Limine and MicroOS, and storage hygiene in the AI coding era: Rust, Docker, model caches, agent worktrees and how to automate cleanup.
LLMs + Agents = competent enough
A synthesis of what I've been measuring and writing: LLMs converged. The transformer architecture got near the ceiling, the parameter war is over, and the leap everyone felt in 2025 and 2026 came from the agentic harness (tool calling, prompt caching, long sessions), not from a base intelligence orders of magnitude larger. I retell the timeline from 2019 onward with sources, explain why the agent allows a less dense model, and why that's already competent enough to automate real software engineering.
Are There Better Languages for Programming with LLMs?
DHH converted his Campfire to several languages with AI and started posting that Rust won. Instead of fighting on X, I built a reproducible benchmark: 7 languages, 4 LLM configurations, 112 implementation attempts, and a queue-under-pressure test with Elixir, Rust, Rails and plain Ruby. Result: no language ranking holds up, concise Ruby cost more tokens than Rust, strong typing spared no repair at all, and under load the four stacks tied on throughput. What matters is testing and discipline, not language.
Which Jev-like AI Decision/Classifier Model Is Best?
I ran my own decision-model benchmark: Jev, Clef, OpenAI Decisions and Microsoft Decision-1 in the cloud, plus eighteen open configurations (Perplexity Decider 27B, Decision 3.0 and Edge, Clef-flash, Kev, Jebadiah, Mapika, Eikos, Laya, Julia, GLiNER, Simple-Jev) on my Strix Halo, across more than 22 thousand classifications with cases in English and Portuguese. The top now has two open-weights models: Perplexity and Decision 3.0 tied at 99.4%, with identical labels on the 1,024 classifications and 100% in Portuguese. Microsoft Decision-1 tied Jev at 97.6% while being the cheapest hosted one. Full table, cost in support emails, gotchas, and where the data converges.
Disaster on My NAS! How I Almost Lost 90 TB of Data
I swapped one drive in my Synology DS1821+ and woke up the next day with two critical drives in an array that only survives losing one. Full post-mortem: why SHR-1 wasn't enough, the warning DSM never sent, the R$ 185 thousand for a QNAP bought in a hurry, and the recovery: parallel NAS-to-NAS rsync, the slow drive holding the array back, the verification of 5.9 million files, the thick provisioning scare, and both NASes healthy in RAID-6.
Understanding the Jev Hype: What Is It For?
Jev is a hosted classifier, and the shock of '193x faster' only exists because so many people use a chat LLM as an expensive classifier. I publish Paulo Câmara's essay (LUA Vision) with his charts, explain what a classifier is and why it's the first thing a computer scientist reaches for, compare the open alternatives Clef and Laya, show the public Decision Index ranking, and close with how to think about this kind of tool in your own project.
September 2026
16 articlesThe Mystery of LUA Vision's Genesys PI - A Brazilian Frontier Model??
After the first Genesys PI test, LUA Vision fixed bugs in the API and I ran everything again: cost dropped 85%, House climbed to 95.5 and Enterprise stayed flat. I also ran a black-box forensic analysis to answer whether the model is a rebadged Qwen, a disguised gpt-oss or an API proxy. I separate what is checked from what is still only their word.
The Limits of LLMs: Good at Reproducing What Exists. Expensive for What Doesn't
Every LLM was trained only on what's public, but 81.5% of GitHub activity happens in private repositories, and the source code of every proprietary software never made it into training. In my retrocomputing projects I felt firsthand where the machine stalls: with no reference and no oracle, it spins in directed brute force.
LLM Benchmark v4: Genesys PI, a New Brazilian Contender!
LUA Vision reached out on LinkedIn offering test access to Genesys PI, a Brazilian model trained from scratch. I ran both tiers, House and Enterprise, through my v4 sabotage benchmark, compared them against Opus, GPT-6, Grok and the Chinese competition, and looked at price, advantages and where it makes sense in the Brazilian market.
LLM Benchmark v4: Opus 5.5, GPT Sol/Luna 6, Mimo 2.6, Grok 4.7
Five new models entered my v4 sabotage benchmark: Opus 5.5, GPT 6 sol and luna, MiMo V2.6 Pro, and Grok 4.7. I compare against external benchmarks, answer whether upgrading is worth it, and Grok 4.7 regresses badly against its own predecessor.
Watch Out! A Pseudo-TechCrunch Tried to Scam Me! And Calendly Is SHIT
An account posing as a TechCrunch journalist DMed me today asking for a meeting, with a Calendly link that redirects to an X OAuth authorization screen. It's the same scam that already caught José Valim, Elixir's creator. I document the scam step by step, with technical proof.
Stop Making Excuses and Ship More! With AI, the Premise Changed. Get It.
With frontier LLMs making fewer mistakes than most human devs, the only real excuse left for holding back a deploy is your own review process's bureaucracy. I use my open source projects, cross-platform CI/CD, and real contribution numbers to show how I've been doing this since February.
New AI-Focused Language Just Released: Bend 2
Bend 2, from Brazilian researcher Victor Taelin, arrived promising C speed, GPU parallelism, and formal proofs that block AI mistakes. I tested it by porting two of my Rust projects right at launch. Plenty of potential; stdlib and ecosystem are still toys.
AI Unmasked the Climate Change Hysteria
AI's race for energy buried the climate agenda: big techs restarting nuclear plants, Germany admitting the mistake, datacenters already consuming double the global EV fleet. With Bjorn Lomborg's numbers, the climate apocalypse as a misallocation of trillions.
Talking a Bit About My AI Skills
Agent skills are just prompts in text files. I keep 25 in my my-skills repo, from pr-audit to fact-check, and they have already closed over a thousand PRs and issues across my projects. The recommendation: don't copy somebody else's skill, build your own.
Why Things Like TypeSafe AI Don't Interest Me
A handful of people showed up in the comments asking 'have you tried TypeSafe?'. I use that as a hook to lay out my rule about new AI tools, actually investigate the case, and explain why I don't waste time optimizing token spend.
New LLM Benchmark v4: Retesting 39 LLMs (Part 2)
The full v4 table, nearly forty models ranked by vigilance, not completeness: six ties at the top, a free model beating half of Claude, GLM winning on a flat-rate plan, and the test I ran to find out whether Kimi or DeepSeek are secretly hiding a connection to Claude under the hood.
New LLM Benchmark v4: Retesting ALL the Top LLMs (Part 1)
I rebuilt my benchmark from scratch for the third time. I threw out an entire version, built a Rails app sabotaged with real CVEs, and spent over $4,000 in 9 days. In this Part 1: the process, the cost, and the why.
Codes of Conduct Are the Root of All Evil: Reclaiming Your Freedom
Codes of conduct were sold as protecting minorities and used to cancel Eich, Stallman, Garfield, and Torvalds. Ten years later, no study shows a diversity gain, and the newest one shows the document pushing out people already there.
You're an Idiot if You Believe the Misleading Ads from OpenAI, Anthropic, NVIDIA. Here's Why
Nvidia and OpenAI declared that AGI has arrived, OpenAI brute-forced a Millennium Problem and threatened a mathematician's career, and an Anthropic employee quit crying Skynet. I take the propaganda apart.
AI Challenge: Converting NES ROMs to the Master System/SMS
I tried using frontier LLMs to convert NES ROMs into Master System games. I explain the architecture, why translating 6502 to Z80 runs slow, the emulator as a feedback oracle, and where the mappers stall everything.
AI-MEMORY 2.0 - The Best Memory System for Agents and Teams
ai-memory hit version 2.0 with the open OKF format, local embeddings on by default, and support for teams working in parallel. I compare it with the competition and show what only ai-memory delivers.
August 2026
11 articlesLLM Benchmarks: The latest Deepseek v4, stop asking
I ran the new Deepseek v4 snapshots through the benchmark: Flash jumped from 80 to 90 and Pro from 82 to 91, both Tier A, the cheapest in the pack. I still prefer Kimi K3 and GLM 5.3, and I explain why a close score does not mean an equivalent model.
Survival Guide for a Censored Internet
From the X block to a sentence enhancement for using a VPN: Brazil is rehearsing real censorship. A three-phase guide (commercial VPN, step-by-step self-hosted VPN, and obfuscated protocols) to keep your communications alive while it is still possible.
Hot Take: Harness, Loop Engineering, Graph Engineering Are Bullshit
I posted that Harness Engineering, Loop Engineering, and Graph Engineering are bullshit to sell consulting hours and courses. Here's my receipt: 30+ public repos, an AI marathon, and a whole benchmark — none of it using any of those 'disciplines'.
Understanding Anthropic's AI Watermark: How to Beat It
Claude now stamps everything it writes with an invisible statistical watermark, a requirement of European AI law. How it works, what the AI Act demands, and the rewrite in another LLM that wipes the signal.
What the Google Quantum Vulnerability Paper Means
Google, the Ethereum Foundation, and Stanford estimate breaking secp256k1 with under 500,000 qubits. I compare GLM, Kimi, and ChatGPT, calibrate the real odds of theft in the coming years, and what you can do today.
LLM Benchmarks: Qwen 3.8, GLM 5.3, Gemini 3.7, Grok 4.6
GLM 5.3 scored 94 and closed in on the leading trio. Qwen 3.8 Max jumped from 51 to 92 after it stopped hallucinating the RubyLLM API. Gemini 3.7 Flash scored 93 — after its first run was voided for cheating. And the local 27B Qwen showed that context is now the bottleneck.
Understanding the Brazilian Censorship of Discord and the Digital ECA Law
Brazil's data protection authority ordered Discord to kill Go Live in the country because encryption prevents moderation. I break down the Naviraí case, the Felca Law, the clash with the LGPD, the new VPN sentence enhancer, and the dangerous precedent it all sets.
Digital David and Goliath: Understanding MegaLag vs Honey/PayPal
Honey spent years skimming creators' commissions with a defeat device written in plain JSON. I walk through the ssd.json line by line, the PayPal-era rewrite, and the class action PayPal failed to kill.
Intel's Comeback: ARM64 vs X86-64 Is Not What You Think
Panther Lake already ties the Apple M5 in multi-core and delivers MacBook-class battery life in an x86 laptop. My thesis: the ISA hasn't defined efficiency since the 90s, and Intel's crisis was managerial, not architectural.
Why a Perfect Digital Election Still Wouldn't Be Viable?
A computer science exercise: how to build a digital election with end-to-end verifiability using commitments, Merkle trees, and zero-knowledge proofs, and why even that 'perfect' system still would not be viable in practice.
Exploiting Coinkite's RNG Egregious Problem
How an attacker enumerates the ColdCard's reduced key space, finds vulnerable wallets on the public blockchain, and moves the funds. Real data from the ongoing theft and step-by-step didactic code.
July 2026
17 articlesURGENT - If You Keep Bitcoin on a ColdCard: MOVE EVERYTHING
An entropy flaw left seeds generated by ColdCard firmware far below the promised security level, with estimated losses above 1,000 BTC. Understand the bug and migrate without repeating the mistake.
Removing DRM from Kindle Ebooks in 2026
I used Calibre, DeDRM 10.0.28, KFX Input, and the Microsoft Store Kindle app in an Omarchy VM to archive 106 books and validate an EPUB conversion workflow I control.
New LLM Benchmark: I Reran Every Test!
I reran the LLM Coding Benchmark with a harder test, three phases, native harnesses, and new tiers. Fable led, Terra tied Sol as the best GPT, and 24 models proved suitable for programming.
AI-Jail: Security Update, Docker Goes Opt-In
Issue #88 proved the Docker socket inside ai-jail gave any agent root on the host. In v1.16.0 the passthrough went opt-in. The flaw, a hands-on demo, best practices, and why Podman was born from this criticism.
LLM Benchmark: Is Opus 5 Any Good?
Opus 5 scored 95/100 on the Rails benchmark, tying Opus 4.8 and edging Fable 5 by one point. Its API rate is half Fable’s, but this narrow test does not define the best LLM.
What's New in My AI-MEMORY: Switch AI Agents Without Losing the Session
ai-memory run keeps the same programming session while switching among Claude Code, Codex, and other harnesses, with searchable workstreams and integration with ai-jail and ai-usagebar.
LLM Benchmark: Should I Use the Highest-Scoring Model?
Why the top entry in an LLM ranking is not necessarily the best model, why 90+ is a cluster, and how to use benchmarks without outsourcing your technical judgment.
LLM Benchmark: Has Kimi K3 Reached Claude Opus Level?
I had Kimi K3 build a chat in Rails 8 on its own: it scored 89/A, beat Opus 4.6, and was a cheaper alternative to Opus 4.8. It still lagged behind in architecture, hardening, and tests.
Quantum News: Majorana 2 and Understanding Shor
Microsoft announced Majorana 2 with 20-second qubits and physicists answering that nothing was resolved. I take the chance to properly explain Shor's algorithm, why factoring becomes period finding, and what quantum computers are actually good at.
Using AI to Solve My Little Day-to-Day Problems
A roundup of my small open source projects: a desk clock with widgets, manga readers, decent email, typing practice, YouTube karaoke, ComfyUI in Docker and more. All born from real little problems in my day-to-day.
How Do I Protect Myself From My Agents Deleting My Stuff?
LLMs deleting files from famous people made headlines this week. In five months of heavy use, in YOLO mode, it never happened to me. But I don't trust them either: BTRFS snapshots, restic backups, sandboxing and discipline.
The Bun in Rust Response Andrew Kelly Should Have Written
I used Claude Fable 5 and GPT 5.6 Sol as judges to compare Rust and Zig in Bun’s rewrite. Rust helps reduce ownership bugs and maintenance costs, but the agent-based method mattered just as much as the language.
LLM Benchmarks: Grok 4.5 and GPT 5.6 Sol
Grok 4.5 entered Tier A with 87, and GPT 5.6 Sol scored 92. Sol beat GPT 5.5 by 92 to 81 in the blind A/B test, but re-auditing brought it down to 85. With cache-read fixed, Opus 4.8 remains the best API balance.
I Had Fable 5 Analyze the Code of TikTok, Clash of Kings and Gov.br - Understanding Fingerprinting
A static analysis with Fable 5 compares TikTok’s opaque fingerprint, Clash of Kings’ persistent UUID, and gov.br’s privacy precautions, with caveats about the method.
Frank GO: Playing Go with AI
I built Frank GO as a Sabaki fork with KataGo, bringing together more than 4,700 tsumego, historical games, and 15 of Hikaru’s Go games. Overlays make territory and scoring visible to beginners.
LLM Benchmark: Sonnet 5 Fails, Gemini Flash Surprises, Sakana Fugu Almost Reaches Tier A
In the coding benchmark, Sonnet 5 scored 58/100 after calling a nonexistent API, while Gemini 3.5 Flash scored 93/100. Sakana Fugu Ultra landed at 79/100, close to Tier A.
Off-Topic: Do Not Put Hope in Politics - How to Become Immune
I argue for treating politics as an environmental risk you manage, like rain and exchange rates, instead of cheering for it like a soccer team. The strategy: grow your income in hard currency, understand taxes, and organize your assets within the law.
June 2026
10 articlesWhy LLMs Will Fail at Your Company
The deliberately speculative thesis is that LLMs amplify processes where people outsource decisions. The proposed alternative is daily cycles of decision, implementation, testing, and review close to the outcome.
ai-memory: long-term memory (Karpathy Wiki) and self-improvement (Hermes) for your projects
ai-memory turns agent sessions into a searchable Markdown wiki with hooks, MCP, and handoffs between tools. A Hermes-inspired loop promotes memories with validation, evidence, and optional review.
The Rio 3.5 LLM Controversy. Plagiarism?
Public evidence indicates that Rio 3.5’s initial checkpoint was a blend of Nex-N2-Pro and Qwen, without credit to Nex. The case points to an attribution failure, not legal plagiarism.
ai-memory: Emergent Architecture and Malleable Software
In 24 days, contributors took ai-memory from a personal MVP to a cross-platform, multi-user system. The architecture emerged from use and was only consolidated afterward.
LLM Benchmark: Kimi v2.7 Code, GLM 5.2, MiniMax M3 Local
In the benchmark, GLM 5.2 scored 87 and Kimi K2.7 Code scored 86. The open MiniMax M3 doesn’t fit in 128 GB, while serious programming still calls for Opus 4.8 or GPT 5.5.
Bypassing the GitHub API Block in Brazil
I tested GitHub’s API block in Brazil: changing DNS didn’t fix it on my connection, but a SOCKS5h wrapper with Tor made gh and ghpending work. It’s a workaround for an opaque block.
LLM Benchmark: Fable 5 and the Anthropic Soap Opera
Fable 5 scored 94/100, practically tied with Opus 4.8, but costs ten times more and redirects sensitive tasks. Since Mythos isn’t auditable, I’m sticking with 4.8.
Playing with TUIs and LLMs - Ratatui+Bubbletea
I built ratatui-bubbletea to give Ratatui Bubble Tea-inspired themes and components using Rust and LLMs. The toolkit already powers real TUIs like ai-usagebar without replacing the base library.
AI Controversy in Open Source Project Contributions - My Take
AI in open source contributions is here to stay, even with AI slop, regressions, and false bugs. The practical answer is to automate triage and auditing without taking the final decision away from humans.
LLM Benchmarks - An Update on Grok 4.3, MiniMax v3, and Opus 4.8
In the Rails 8 benchmark, Opus 4.8 keeps the lead at 95/100, while Grok 4.3 and MiniMax M3 finally become usable but remain in Tier B, behind GPT and Opus.
May 2026
12 articlesOpen Source Best Practices with LLMs - The Bare Minimum
Open source projects built with LLMs are only ready with simple installation, reliable tests and CI, and documentation focused on the problem. Standardizing releases and deployments makes automation predictable.
Complete Manga Solution: Frank Manga+, Frank Yomik, and the Prettify-Manga Extension
Frank Manga+ brings the MANGA Plus catalog to the desktop, while Prettify improves reading on websites and Kindle, and Frank Yomik adds furigana or translation to Japanese and Korean pages.
Backing Up Gmail to Maildir on Linux
I replaced manual Thunderbird backups with mbsync in pull mode, Maildir, and a systemd timer. Restic replicates the archive to NAS and off-site storage, keeping emails local, open, and accessible to multiple clients.
First Impressions Using Oh-My-Pi and OpenCode
Oh-My-Pi is more flexible for projects with varied files and artifacts, while OpenCode is more polished for ordinary code. The test showed that no harness audits everything, and Codex remains behind.
Akita's AI Tips and Toolkit: ai-jail, ai-memory, ai-usagebar
After more than 600 hours with coding agents, Akita brings together tools like ai-jail, ai-memory, ai-usagebar, and ghpending, and argues for testing, backups, refactoring, and disciplined engineering.
I Built a Waybar Widget for Omarchy to Monitor LLM Plan Usage: ai-usagebar
ai-usagebar, a Rust port of claudebar, brings Claude, Codex, Z.AI, and OpenRouter into a Waybar widget or standalone TUI that works in any terminal, including over SSH.
I Built a CLI to Check My GitHub Pending Stuff: ghpending
Akita built ghpending to query issues and PRs across multiple repositories in a Rust digest, fetching everything in parallel and raising GitHub’s limit from 60 to 5,000 requests per hour with a token.
I Built a Memory System for Coding Agents: ai-memory
After finding reindexing, data loss, and broken hooks in agentmemory, Akita built ai-memory in Rust with Markdown, SQLite/FTS5, automatic capture, and agent handoff. It’s still beta.
AI Agent Memory: Karpathy LLM Wiki and agentmemory in Practice
Akita compares compaction in Claude Code, Codex, and opencode, explains Karpathy’s LLM Wiki, then tests agentmemory before dropping it over structural bugs and pointing to ai-memory.
Wrapping Up My AI Marathon: Success or Failure?
After more than 500 hours across 24 repositories, Akita ranks his experiments and concludes Codex and Claude Code support daily coding, but without review and XP, productivity turns into slop.
LLM Benchmarks: DeepSeek Unlocked! Use DeepClaude
I swapped opencode’s harness for Claude Code’s loop through DeepClaude, and DeepSeek V4 Pro rose from 69/B to 89/A in 18 minutes and $3.14. The controller’s multi-turn bug remains.
NW-Omarchy: Bringing Omarchy to X11 with XLibre
I built NW-Omarchy as an X11 session alongside Omarchy, using XLibre, bspwm, and equivalents to preserve themes and 70-plus shortcuts. HDR and fractional scaling stay in Hyprland.
April 2026
18 articlesLLM Benchmarks: Is It Worth ($$) Mixing 2 Models? (Planner + Executor)
Three rounds show multi-agent doesn’t beat solo Opus 4.7 in opencode, which scored 97/100 in 18 minutes for about $4. GPT 5.4 xHigh with a medium executor saves money but loses 3 points.
LLM Coding Benchmark (May 2026): DeepSeek v4, Kimi v2.6, Grok 4.3, GPT 5.5
I reaudited 24 LLMs on the same Rails app with RubyLLM: Opus 4.7 and GPT 5.4 tie at 97, GPT 5.5 reaches 96, Kimi and Gemini become Tier A, and DeepSeek V4 Pro reaches Tier A only through DeepClaude.
How DriveClub and shadPS4 Almost Defeated AI and Me: How to Learn
After 31 phases and 44 commits, I found DriveClub’s nighttime blackout in shadPS4 came from unsynchronized GPU→CPU feedback. `readbacks_mode: 2` fixed the bug, but the investigation taught me the path.
Clean Code for AI Agents
Clean Code carries different weight when an agent is the primary reader: small code, greppable names, provenance context, headless tests, and explicit rules reduce navigation, cost, and errors.
My Favorite Retro Racing Games Running on My Distrobox
I consolidated a collection of classic racing games on Linux in Distrobox. Driveclub runs at 30 FPS with a shadPS4 fork, and FM4 works in Xenia after extensive tweaking.
LLM Benchmarks Part 2: Is It Worth Combining Multiple Models in the Same Project? Claude + GLM??
I tested seven combinations in Claude Code, opencode, and Codex, but no sub-agent was called. Variance across runs and harnesses keeps solo Opus as the default for greenfield projects.
Omarchy on the Thinkpad T14 Gen 6: Mini-Review and Full Setup
I installed Omarchy on a ThinkPad T14 Gen 6 and turned the laptop into a Linux companion for SSH, Claude Code, and NAS. The screen feels dated, but ports, durability, and support make up for it.
Why LLMs Aren't Giving You the Result You Expect | Why I Prefer Claude Code Today
Every week somebody tells me 'I canceled my Claude plan, it just doesn't perform as well as GPT for me.' I've got 500+ hours in Claude Code and Codex, 400k lines generated, and both deliver. The difference isn't the model. It's how people are talking to it.
Seedance 2.0 Is Finally Out: First Impressions
ByteDance opened Seedance 2.0 to the public today after months of restricted access. I tested audio-driven lip sync with my anime avatar, and a Blender render fed in as a video reference. There's real work you can do here, but it's nowhere near professional production yet. Also: deepfakes just stopped being hypothetical.
An Emulation Distrobox with Claude Code
I built an Arch Distrobox for 12 emulation platforms and Steam, organized into 17 Ansible roles with Claude Code’s help. The automation makes setup reproducible and keeps decisions visible.
VS Code Is the New Punch Card
In the era of coding agents, typing everything by hand in VS Code is becoming the modern equivalent of a punch card. Foundations did not become legacy.
How ElevenLabs Was Not Killed by Qwen3 TTS
When Qwen3 TTS dropped, half the internet called it an 'ElevenLabs killer'. I spent weeks trying to run Qwen3 in production on my podcast. Yesterday I finally switched to ElevenLabs v3. Less than a day later, I can tell you: open source is still miles behind.
20 Years of Blogging: Translating Everything to English
Four days ago I hit 20 years of blogging. When I sat down to write the anniversary post, I ended up doing something I never had the bandwidth for: translating the whole blog to English. With Claude Code, over a weekend.
Is RAG Dead? Long Context, Grep, and the End of the Mandatory Vector DB
Frontier models went from 200k to 1M tokens of context. Does it still make sense to wire up a whole vector DB stack to do RAG, when you can just grep and dump the entire document into the window?
Testing Open Source and Commercial LLMs - Can Anyone Beat Claude Opus?
This historical benchmark compared 33 LLMs on a Rails app: Opus, Sonnet, and GLM 5 worked, while Qwen 3.6 35B came close on an RTX 5090 after a fix. The rankings were later revised.
Turning YouTube into a Karaoke App | Frank Karaoke
I built a Flutter Android app that overlays live scores on YouTube karaoke videos, using audio filtering and YIN in four modes. It’s fun, but scoring is still experimental.
Bitcoin on the Home Server: Sovereignty and Privacy with Coldcard, Sparrow and Fulcrum
I built a home Bitcoin stack with Coldcard, Sparrow, Fulcrum, and bitcoind: offline keys, private queries, and my own broadcast. It takes work, but gives me more control over custody and operations.
My Sim Racing Cockpit - Formula FX1
I replaced years of unstable stands with a Formula FX1 cockpit featuring Fanatec direct drive, PS5, RTX 4090, and a 4K OLED. It doesn’t wobble and is ready in 30 seconds for my single-player use.
March 2026
13 articlesClaude Code's Source Code Leaked. Here's What We Found Inside.
I analyzed Claude Code’s leaked source map and found hidden features, layered memory, multi-agents, and DRM based on xxHash64. The code also exposed a product that’s difficult to maintain.
Migrating my Home Server with Claude Code | openSUSE MicroOS
I migrated 49 containers from Ubuntu to openSUSE MicroOS with Claude Code’s help, reorganizing Docker, NFS, ROCm, backups, and SELinux. Automation sped things up, but architecture and validation stayed human.
Review: Minisforum MS-S1 Max | AMD AI Max+ 395 with 96GB of VRAM
In my tests, the RTX 5090 was up to 7 times faster on models that fit within 32GB. The Minisforum, however, ran 50-to-81GB models the card can’t handle.
Teaching People to Question the News | Frank Investigator
I built Frank Investigator to compare coverage, track omitted sources, and analyze news in 15 stages with three LLMs. The report flags omissions and reframing, but doesn’t issue final verdicts.
I Rewrote OpenClaw in Rust. Did It Work? | FrankClaw
I rewrote OpenClaw’s core in Rust as FrankClaw, with 56,586 lines, 7 channels, and fixes for the 7 audited critical vulnerabilities. It handles simple conversations, but complex workflows remain untested.
Going After Email Fraud | Frank FBI
I built Frank FBI to analyze forwarded emails in six layers, combining authentication, reputation, OSINT, and three LLMs. The result is a self-hosted risk report, with the data under your control.
Porting 10K Lines of Python to Crystal with Claude: easy-subtitle
I asked Claude to port Subservient to Crystal with feature parity and got easy-subtitle in under 40 minutes: 2,516 lines, 76 tests, and a 6 MB static binary with no Python runtime.
Crystal and a Smart FFmpeg Wrapper Built in 3 Hours | easy-ffmpeg
I wanted to convert videos without memorizing FFmpeg flags. Out came a smart CLI in Crystal with presets, interactive mode and static compilation for Linux and macOS, in 3 hours of vibe coding.
37 Days of Vibe Coding Immersion: Conclusions on Business Models
After 37 days, 650+ commits, ~144K lines of code and nearly 10 published projects, my conclusion on what vibe coding means for the future of startups and business models.
My First Vibe Code Failure and How I Fixed It | Frank Yomik
How I spent days building a manga speech bubble detection system with OpenCV, threw it all out and rebuilt it in hours with a pretrained model. Lessons on vibe coding, productive failure, and knowing when to change course.
I Built a Data Mining System for My Influencer Girlfriend — Tips and Tricks
How I built a Rails 8 + SQLite + Discord bot data mining system for my influencer girlfriend in 3 days, 58 commits, with 40+ tools accessible via LLM tool calling.
ai-jail: Sandbox for AI Agents — From Shell Script to Real Tool
ai-jail is a Rust tool that wraps bubblewrap (Linux) and sandbox-exec (macOS) to safely run AI coding agents like Claude Code, Codex, OpenCode and Crush in a sandbox.
Software Is Never 'Done' — 4 Projects, Life After Deploy, and Why One-Shot Prompting Is a Myth
125 post-production commits across 4 projects in 10 days. Real bugs, real users, real iteration. Why one-shot prompting is a myth and software is never 'done'.
February 2026
17 articlesRANT: Did Akita Bend Over for AI??
I have been 'bending over' for AI since 2023. If you only watch out-of-context podcast clips, let me spell it out for you.
Vibe Code: I Built a Smart Image Indexer with AI in 2 Days | Frank Sherlock
How I built Frank Sherlock, a local AI-powered image catalog desktop app, in a weekend using Agile Vibe Coding with Claude Code.
Vibe Code: I Built a Mega Clone in Rails in 1 Day for My Home Server
Building FrankMega, a self-hosted Mega.nz clone in Rails 8, in a single day with Claude Code, and why experience still matters in vibe coding.
From Zero to Post-Production in 1 Week - How to Use AI on Real Projects | Behind The M.Akita Chronicles
How 274 commits in 8 days shipped a full newsletter, podcast, blog, and Discord bot to production using Extreme Programming with Claude Code as the pair.
Integration Tests in a Monorepo | Behind The M.Akita Chronicles
How a monorepo running three apps that share a filesystem uses a dedicated integration environment, DevCache, rsynced production data, and preflight checks to catch the bugs unit tests never see.
SQLite + Kamal: Rails Deploy Without Drama | Behind The M.Akita Chronicles
How Rails 8 with SQLite and Kamal lets you run a full production app on a $12/month VPS with zero external services.
Discord as an Admin Panel | Behind The M.Akita Chronicles
How I replaced the classic Rails admin panel with a Discord bot: parser/dispatcher patterns, embeds, reactions, tool-calling LLMs, and the lessons from failing silently in production.
Async Jobs That Survive Chaos | Behind the Scenes of The M.Akita Chronicles
Production-grade patterns for Rails 8 ActiveJob and SolidQueue: retries, distributed locks, atomic email claiming, orchestrators, cron safety nets, and status notifications.
Frontend Without a Framework | Behind the Scenes of The M.Akita Chronicles
How I built a modern, responsive, dark-mode, email-compatible, multi-platform frontend for The M.Akita Chronicles without a single line of JavaScript framework.
Serving AI in the Cloud: My Personal TTS | Behind the Scenes of The M.Akita Chronicles
Behind the scenes of putting Qwen3-TTS in production on a serverless GPU to auto-generate a weekly podcast: cold starts, voice cloning, sampling, loudness, and prompt engineering.
Web Scraping in 2026 | Behind the Scenes of The M.Akita Chronicles
How I went from a four-line HTTP GET to packing a full Chromium into Docker just to read the news for my newsletter in 2026.
Sending Emails Without Getting Flagged as Spam | Behind The M.Akita Chronicles
How to build a reliable newsletter sending pipeline on Amazon SES, covering atomic claiming, terminal states, DKIM/SPF/DMARC, List-Unsubscribe and the silent SES suppression list.
Vibe Code: From Zero to Production in 6 DAYS | The M.Akita Chronicles
How I shipped a full newsletter, blog, podcast, and Discord bot to production in six days using vibe coding the right way.
AI Agents: What Would Be the Best Programming Language for LLMs?
A thought experiment on what a programming language designed for LLM coding agents, rather than human programmers, would look like.
RANT: Did AI Kill Programmers?
Akita's rant on why AI won't replace real programmers, why the programming bubble already popped, the S-curve ceiling of current LLM architecture, and what actually changes for engineers in the LLM era.
Vibe Code: I Built a Markdown Editor From Scratch With Claude Code (FrankMD) PART 2
Part 2 of the FrankMD story: refactoring, i18n, syntax highlighting, tests, and the hard-earned lessons from 30 hours of vibe coding with Claude.
Vibe Code: I Built a Markdown Editor From Scratch With Claude Code (FrankMD) PART 1
How I vibe-coded FrankMD, a self-hosted Markdown editor tailored to my blogging workflow, in three days with Claude Code.
January 2026
10 articlesVibe Code: Which LLM Is the BEST?? Let's Talk for REAL
I ran the same extensive analysis pass through Opus 4.5, GPT 5.2 Codex, Kimi 2.5, Gemini 3 Pro and MiniMax v2.1 on a tiny app, and every single one still left things on the table.
Vibe Code: I Built a Little App 100% with GLM 4.7 (TV Clipboard)
A full walkthrough of vibe coding a tiny Go WebSocket app with GLM 4.7, showing in practice what the 80/20 rule really costs in time, tokens and intervention.
AI Agents: Which One Is Best? OpenCode, Crush, Claude Code, GPT Codex, Copilot, Cursor, Windsurf, Antigravity?
A practical take on the current state of AI coding agents, why proprietary harnesses matter, and why I am sticking with Crush.
AI 3D: Can You Actually Model 3D With Prompts Now?
A practical walkthrough of generating high-quality 3D models from prompts and images using Nano Banana, Hunyuan, Hitem3D, and Blender, all the way to a physical Bambulab print.
AI Agents: Is GLM 4.7 Flash really that good?
Running GLM 4.7 Flash locally on an RTX 5090 via Ollama to see if it can finally match commercial LLMs on a hard coding challenge.
Omarchy 3: Dual GPU Setup with AMD and NVIDIA
How I configured Hyprland on Omarchy to use an AMD iGPU as primary and offload 3D apps to an NVIDIA RTX 5090 via Prime.
AI Agents: Installing LSPs for Crush
My Crush config with LSPs wired in, plus an Arch Linux script that auto-detects installed languages and installs the matching language servers.
AI Agents: Comparing the Top LLMs of 2026 on the Zig Challenge
I put the top commercial and open source LLMs of early 2026 through a nasty Zig + llama.cpp migration challenge to see which ones actually deliver.
AI Agents: Locking Down Your System
I show how to isolate coding agents with Bubblewrap: the jail leaves only the project writable, hides the rest of the system, and mounts only what is needed, reducing the risk of destructive commands.
Omarchy 3 - One of the Best Coding Agents Out There: Crush
I used Crush with Claude Opus 4.5 on a Hugo blog and a Zig chat using llama.cpp, CUDA, and Qwen3. It refactored scripts, updated APIs, fixed a segmentation fault, and delivered a working binary for nearly $8.
September 2025
12 articlesOmarchy 2.0 - LazyVim Basics
A hands-on introduction to LazyVim for people coming from VS Code or Sublime Text, covering the essential Vim navigation, editing, and block commands to get productive fast.
Omarchy 2.0 - Recommended for Beginners?
Why Omarchy is my daily driver, why there is no such thing as a "distro for beginners", and how to use Ventoy and Distrobox to test distros yourself instead of outsourcing your decisions.
Omarchy 2.0 - Install with the Omarchy ISO
Why the official Omarchy ISO is the easiest way to install Omarchy, with LUKS encryption, Limine+Snapper rollback, and Hyprland+UWSM auto-login out of the box.
Omarchy 2.0 - Bitwarden Self-Hosted / VaultWarden
Self-hosting a Bitwarden-compatible password manager with VaultWarden on a home server, routed through a Cloudflare tunnel, with 2FA via Aegis.
Installing Grafana on My Home Server
How I set up Grafana, Prometheus, and cAdvisor with Docker Compose to monitor my Intel NUC home server.
Protecting Your Home Server with Cloudflare Zero Trust
How I added a Cloudflare Zero Trust login layer on top of my home server services, using Google OAuth as the identity provider.
Accessing My Home Server With a Real Domain
An advanced guide to exposing a Docker-based home server through Cloudflare Tunnels with a real domain and proper HTTPS.
Omarchy 2.0 - Understanding SSH and Yubikeys
A practical guide to SSH keys, Yubikey 5 hardware-backed keys, and ssh-agent on Arch Linux and Omarchy.
Omarchy 2.0 - TUIs (Terminal User Interface Apps)
Why TUIs deserve a spot on your desktop in 2025, plus a tour of my favorite terminal apps and how Omarchy makes them feel native.
Omarchy 2.0 - Mise for Organizing Development Environments
How to use Mise on Omarchy (or any Linux distro) to pin language versions per project and stop breaking your apps on every system update.
Omarchy 2.0 - LazyVim - LazyExtras
A practical walkthrough of LazyVim and the :LazyExtras command that ships pre-installed with Omarchy, covering the basic hotkeys and how to enable language support.
Omarchy 2.0 - ZSH Configs
How I swapped Omarchy's default Bash for ZSH and wired up Atuin, Starship, secrets, and handy aliases to match my workflow.
August 2025
2 articlesHow to Contribute to the AkitaOnRails Blog Using Docker
A quick walkthrough on using Docker to spin up the AkitaOnRails blog development environment and start contributing without installing Hugo, Go, or Ruby locally.
Installing Omarchy 2.0 from Scratch - Personal Notes
My personal notes walking through a clean install of Arch Linux with Omarchy 2.0, covering BTRFS, Timeshift snapshots, separate HOME, NFS, and Hyprland customization.
June 2025
1 articleAGI or Skynet Isn't Coming Anytime Soon
Why the current LLM architecture cannot reach AGI, and why the hype from Big Tech CEOs and media-anointed experts is mostly backroom negotiation noise.
May 2025
7 articlesYour Windows May Be Crippled Without You Knowing. Check This!!
How a hidden Windows Power Saver profile was silently throttling a high-end desktop to a fraction of its real speed, and how to fix it.
Computing History and Retro Dev on YouTube
A curated list of YouTube channels on computing history and retro development that every dev should watch.
Final Attempt to Train an LLM with LoRA. Cannon Shot, But Missing the Fly.
Renting an 80GB H100 on RunPod to fine-tune Qwen3-32B with a LoRA on Zig 0.14, and discovering that even cannon-class hardware can still miss the target.
Teaching the Latest Zig to Your LLM - Training LoRAs (Sort Of)
A hands-on walkthrough of training a LoRA adapter on Qwen3-8B to teach it about Zig 0.14, covering datasets, fine-tuning, vLLM serving, and the limits of the experiment.
RANT - LLMs are LOOT BOXES!
Why LLMs for real coding behave like gacha loot boxes, and why every incentive in the industry pushes you to burn more tokens.
When Do LLMs Fail at Programming? A More Realistic Use Case.
A real-world attempt to build a Zig program using llama.cpp with Gemini 2.5 Pro and Aider, showing where LLMs crumble on new languages and libraries.
Rant - Will LLMs Evolve Forever? Demystifying LLMs in Programming
A hype-busting rant on LLM benchmarks, the myth of exponential AI progress, and my hands-on experience showing why Vibe Coding does not work.
April 2025
17 articlesDissecting an Ollama Modelfile - Tuning Qwen3 for Code
A walkthrough of sampling parameters (temperature, top_p, top_k, min_p, repeat_penalty) and how to build a custom Ollama Modelfile to tune Qwen3 for software development tasks.
Testing the Newly Released Open Source LLM - Qwen3 (with Aider and Ollama)
First hands-on impressions of Qwen3 running locally via Ollama and remotely via OpenRouter, tested with Aider on real refactoring tasks.
Destroying the ChatGPT 4o "Personality"
How to disable ChatGPT's pseudo-human chatter and get straight, dry technical answers by tweaking the custom traits settings.
Testing LLMs with Aider on RunPod - which one to use for code?
Hands-on testing of Qwen2.5 Coder, DeepSeek, Codestral and others with Aider, locally on an RTX 4090 and on RunPod, comparing them against Claude and Gemini.
Your Own Free Universal Co-Pilot Running Local: AIDER-OLLAMA-QWEN
How to run Aider with Ollama and Qwen 2.5 Coder locally to build a free, universal Co-Pilot alternative that works with any editor.
LLM Hello World: Building Your Own Local AI Chat
A hands-on Hello World experiment explaining how LLMs actually work under the hood, building a tiny local chat CLI with Qwen 2.5 Coder running on a single GPU.
Accessing Your NAS Using iSCSI Instead of SMB
Why I moved my Docker data-root off NFS and onto an iSCSI virtual drive served from my Synology NAS, and how block-level storage crushes file-based protocols for high-churn workloads.
Changing Clothes Using A.I. (ComfyUI)
A walkthrough of a ComfyUI workflow using IDM-VTON and IPAdapter to swap clothing on photos while preserving the original face and pose.
Using A.I. (ComfyUI) to Generate NPCs in Game Development
A walkthrough of Mickmumpitz's ComfyUI workflow for generating coherent NPC character sheets from scratch, running on my Docker Compose setup.
Understanding the Basics of ComfyUI to Generate AI Images
A hands-on, no-nonsense walkthrough of ComfyUI's core concepts (checkpoints, LoRAs, ControlNet, VAEs, samplers) and a real workflow for turning a photo into anime-style art.
Generating AI Images - even Ghibli style 😂 - with Docker and CUDA
A technical walkthrough of running ComfyUI locally with Docker and CUDA to generate AI images, including anime and Ghibli styles, with full control over models, extensions, and workflows.
Generating Up to 2-Minute Videos from a Photo with A.I.
Using FramePack with Tencent's HunyuanVideo in Docker on an RTX 4090 to animate photos of action figures and drawings into videos up to two minutes long.
Upscaling Old Anime to 4K with AI
Using Real-ESRGAN and Video2X inside a CUDA Docker container to upscale old anime files to 4K on an NVIDIA RTX 4090.
Colorizing Black and White Images with A.I.
Using DDColor inside Docker to colorize old black and white photos, plus a Reinhard global color transfer hack to bias the result toward a reference image.
Configuring My Synology NAS with NFS on Linux
How I switched from SMB to NFS to mount my Synology DS1821+ shares on Linux with proper ownership and permissions, plus the UID/GID gotcha along the way.
NVIDIA and Wayland - Problems for PCI Passthrough in VMs
How GNOME on Wayland breaks live NVIDIA GPU passthrough to QEMU/Libvirt VMs, and the workaround scripts I use to make it work again.
BIOS Configuration of My PC - X670E Aorus Xtreme
My BIOS tuning notes for a Ryzen 9 7950X3D on a Gigabyte X670E Aorus Xtreme, plus Linux-Zen kernel and an automated BIOS update checker.