Have become top 3 trending developer on Github, with @mitchellh right behind me 😱

I love how agents do the estimations - it will take a 2 days of work - and then complete it in 15 minutes.

Some of us will get 100% increase on the AI bills starting from new year. Read the small text. Google going to 2x price for all its models soon. Below the prices for Gemini 3.8 Flash model (which I love btw), but it makes a lot of my use-cases simply not economically feasible. Miss the Gemini 2.5-flash, it was amazing 🥲

Recording of my presentation on verifying AI generated code piped.video/watch?v=Qh5SGXSs…

What I'm building: Spirit Room, a test environment for MCP workflows. Start with a user job and an expected result. Run it across agent apps. Inspect where it breaks. Test the fix. Currently a hands-on pilot. What job should your MCP pass reliably? spiritroom.app/

Today I'm going to present at the @techexcellence_ Meetup on the topic of verifying AI-generated code and how we can go from AI-assisted to agentic development. It is free, and a recording will be available. meetup.com/ru-RU/techexcelle…

Working on a new visualisation for my software factory, what do you think?

A passing MCP tool test doesn't prove the user's job gets done. Maintain tests with real prompts and expected outcomes. Inspect searches, tool choices, arguments, retries and final answers. Keep failures as regression cases. Find the gaps in your MCP: spiritroom.app/#contact

One MCP list response we captured was over 46,000 tokens. The data can be correct and still be hard for an agent to use. Page the results. Add filters. Put the summary and cursor first. Then inspect what actually reaches the model. Raw JSON size doesn't tell the whole story.

Every MCP output shapes the agent's next decision. "Invalid ID" leaves it guessing. "Group not found. Call groups_list, select the matching group, then retry with its ID" gives it a recovery path. Success needs direction too: what happened, what's next, and when to stop.

Your MCP description might be 2,000 characters long. The agent might initially see only the name or a short preview. Put the action, user vocabulary and critical prerequisite first. "Email an invoice (bill). Requires an invoice ID." Make the opening earn the call.

Started experiment with AI generated videos, on auto-pilot, like this piped.video/shorts/4cv7mWTqs… Have learned so much in the the process: automatically building the scripts, hooks, montage, various effects, speech with emotions, and etc. Running experiments on various axis right now to find the format which works the best. Highly recommend to try by yourself. All running through Grok Bot 👌

So much ideas i have depend on the time when we have LLMs with such TPS! Look forward to testing it!

So basically, today PHP got two critical bugs that let anyone run code on your server without even logging in, which is bad, and then the details leaked before the patch was even out, which is worse, and it works on WordPress and MediaWiki and Magento and Joomla and Laravel, and Censys says like 1.5 million servers are running a version with no fix, so the NSA was like "take your system offline," and the FBI was like "take your system offline," and CISA and the UK and CERT-EU were all like "take your system offline," and then the Prime Minister of Japan, the actual Prime Minister, posted at 5:55 in the morning telling 48.8 million people to take their system offline, so everyone took their system offline, like the Dutch cybersecurity agency took its own website offline, and the Dutch parliament went offline, and the Port of Rotterdam went offline, and Finland and France and Denmark started pulling government sites, and Wikipedia went down, and Archive.org went down, and Asana and GOG and Something Awful went down, and the University of Vienna's webmail went down for some reason, and that's just PHP. Because GitLab also dropped a CVSS 10.0, which is the highest score they give out, for a bug that's been there since 2014, so that's every self-hosted GitLab for twelve years, and then LiteLLM 1.107.0 got hijacked and shipped a credential stealer and a remote access trojan in every single wheel and the Docker images, including latest, and you didn't even have to import it, just installing it was enough, and it's the second time this year, and then Railway paused every deployment on the entire platform because someone got into the environment variables, so now everyone has to rotate every secret they've ever stored there, and TiDB Cloud said someone was walking around inside with stolen credentials for basically a month and an extortion group says they took 10.6 terabytes, and Zeabur got breached for the second time in two months and is just shutting down, like the whole company, and Node.js patched remote code execution and a TLS bypass, and CISA says Plex is being actively exploited, and someone took over 300 Tor exit relays to steal SSH passwords, and SafePay ransomware has hit over 600 critical infrastructure orgs, and India's biggest bank filed a data loss incident, and Japan has been in an official state of emergency in cyberspace since October 9, which honestly explains the 5:55 AM post. Oh, and eMudhra, a certificate authority that's issued over 60 million certificates, whose entire job is proving websites are who they say they are, was supposed to be using private keys that are random numbers over a hundred digits long, and one of them was 4. Just 4. And the others were like 66 and 237. Anyway, it's Wednesday.

Going to a be a speaker at Tech Excellence online meetup next week! In this session we going to answer on main question of software engineering - how much tests is enough to prove that given feature or bug-fix is working. We prove it mathematically and then turn it to deterministic guardrails which agent can't change, and has to follow. Link below👇

Those who updated to a new SuperGrok billing - does the cursor and grok bot now a unified token counter ? One of the benefits of the old plan was that you essentially got 2 subscriptions for the price of one - Cursor and Grok one. So I wonder if I should delay upgrade.

While we have so many models out of there, only a few popup up - which are smart enough for agentic flows, less then $0.5 per output, and have providers with zero retention policies. Literally just 2 models so far. And openrouter is amazing for doing this kind of research.