Principal Engineer building in public (@onehausapp / @mcpscores). Previously CTO @ Illuvium, Principal @ Checkout dot com, Lead Architect @ ClearBank
London, England
Joined July 2016
- Tweets3.9K
- Following1.5K
- Followers1.3K
- Likes2.7K
CodeWithStu retweeted
ICYMI: For the 10th anniversary of "Site Reliability Engineering", we’ve rewritten the definitive Google SRE book! 📖
Featuring brand-new guidance on SRE for the agentic era, modern SLOs, and culture.
Get your copy today: goo.gle/SREbook
CodeWithStu retweeted
New on our Engineering blog: Creating in-house source control.
In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.
Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: anthropic.com/engineering/re…
Here’s the approach i’m using to secure my agents:
Create an MCP server that only allows specific actions. Connect the agent to that. Then agent can only do what you explicitly allow. Expand the tool list as needed when you need it :)
Worth trying?
Introducing Microsoft-Decision-1
Our model for fast decision-making that delivers top performance in latency and quality on structured decision tasks to outperform both LLMs and other decision models.
commandline.microsoft.com/mi…
honestly nightly is pretty damn stable. you probably won’t notice a stability difference. great job on the ci for that theo + team
didn’t check. But what i did do is have claude write new subagents based on my usage and gave it the prompting guides, generally updating the stack. Also got it to write hooks to blanket stop certain actions like test runs in subagents
This looks incredibly useful
llmfit 1.1.17 is out. Mouse support in the TUI, --memory-percent/--ram-percent flags, more honest speed estimates for CPU offload, deterministic
ranking and 40% faster search.
Which LLMs will actually run on your machine, now with better numbers.
github.com/AlexsJones/llmfit
Yes of course i know :) the big problem is if you don’t understand what you’re asking for, not what’s generated
i’m 90% sure it’s a bug because it’s fine most of the time then all of a sudden you’ve used a ton of usage for no apparent reason :(
Really starting to like the look of this next update. Onboarding experience is going to be so much nicer, plus the new homepage experience. Going to run with experimental feature flag so we can remove it/change it and force people to the old version if any issues.
What'd you think?
CodeWithStu retweeted
Connector submissions for @Muse are open 🚀
If you have MCP go here -> muse.ai/platform and submit ASAP
