Doing things with code; self identified geek; vaccinated; former Snowflake in ЗСУ; @everrunshq; bashkit; yolop
Austin, TX
Joined July 2008
- Tweets31.4K
- Following726
- Followers1.4K
- Likes16.8K
Bashkit v0.20.2 is out! 🐚
Mostly maintaianance release
- Bump for Wasmtime 48.0.8 security fixes
- Signed Windows Python wheels
- Fixes for binary HTTP output and directory copies
On crates.io, PyPI & npm:
github.com/everruns/bashkit/…
Mykhailo Chalyi retweeted
October newsletter!
New durable execution on the actor model for Agents, the Agent API, coordinator mode, and self-hosting with one command or a Helm chart, and much more!
everruns.com/blog/whats-new-…
Mykhailo Chalyi retweeted
Всі люди, які хотіли виїхати з Олешок, вже в дорозі, — Лубінець.
Евакуйовані мешканці вже перебувають у дорозі. Тим, хто їхав з тваринами, також була надана така можливість. До міста також доставили гуманітарну допомогу.
«До нас підходили люди, говорили, що хочуть залишитись тут. Така гуманітарна допомога буде їм додатковою підтримкою», — зазначив Дмитро Лубінець.
На місці перебувала знімальна група телеканалу «1+1». За їх словами, кількість охочих евакуюватися зросла, коли люди дізналися, що до Олешок приїдуть представники України — місцеві боялися російської фільтрації.
російські воєнкори писали, що до України виїхали 17 автобусів, до РФ — лише чотири.
rc14 of pypi.org/project/mistralai-a… have bashkit extra! If it lands in release, that would be next level of awesome!
Mykhailo Chalyi retweeted
Postgres was built to run for years.
But AI agents need databases for 2 minutes!
🚀 Create → Migrate → Test → Destroy 💥
Now multiply that by 1,000 agents running in parallel.
You don't need 1,000 Postgres servers.
You need PGX. 🐘
One runtime. Thousands of isolated databases. Create them in milliseconds. Throw them away when done.
nitter.gq/alxshp/status/21125799…
🚨Introducing a new database: PGX 🐘
Postgres for AI agents. In Rust. pgx.dev
One runtime. Thousands of Postgres databases.
⚡ Up to 100× less idle memory
⚡ 1,000 databases in just 659 MB RAM
⚡ Clone a 674 MB database in 85 ms
⚡ ~0.7 MB per idle database
RUST. OPEN SOURCE. FREE.
No containers per database.
No heavyweight instances.
No production baggage.
Real Postgres. Built to be disposable.
nitter.gq/alxshp/status/21100494…
On the same tenant UI source, a full non-incremental check took 62.6 seconds with tsc and 0.81 seconds with tsrs; both passed.
WTH... I need to check if tsrs is actually doing anything.
Hm, Cloudflare Artifacts:
git push: 1.3 to 1.6 s each. The first push took 2.4 s, and a full clone took 1.7 s.
This is awesome. AWS should be next.
The remote MCP server for Terraform is now in public preview. 🙌
Connect AI agents to HCP Terraform without requiring developers to install or maintain an MCP server locally. This makes it easier to bring governed infrastructure workflows into AI-powered development experiences. 🔌
Learn what’s new for agentic infrastructure lifecycle management. ibm.co/6018EV9h0
Busy release of Everruns! Enjoy!
everruns.com/docs/changelog/…
Everruns 0.49.0 is out 🧵
Top:
- MCP servers with thousands of tools
- MCP Tasks
- Two-way GitHub sync for Agent source
- New Knowledge
- Automatic session titles
- Improvements for self-hosted
- Plus 50+ fixes and improvements
Full list:
everruns.com/docs/changelog/…
But also fun, two fleets of workers were happy claiming work. And CP was okay with that. From user point of view, everything was totally fine, and I guees nobody noticed external traffic was going from different IPs.
Lol, was doing regular load testing exercise for Everruns, and had my outside of the zone stendby running. This fooker claimed tons of background work, resulting in 650 GB/day API traffic for last couple of days :D
AWS released runtime that can host 75k agents in 8GB. I guess I need to run this test for Everruns.
aws.amazon.com/blogs/opensou…
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected!
There was no guardrails blocking each and other request!
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
