AI Engineer building agent tools and workflows @WorkOS https://xitter.cf/t.co/FyPcujTvkz

Idaho
Joined December 2024
I built pi-posthorse, a Pi extension inspired by Codex's new approach to context management. Instead of relying on compaction, agents can inspect context usage, reset their context window, write durable notes, and search their full session history. github.com/fitchmultz/pi-pos…
28
42
8
874
123,380
Apparently my ChatGPT account is too secure to connect GitHub 😂 Hey @OpenAIDevs, the GitHub connector requires MFA, but my ChatGPT account has Advanced Account Security enabled with 6 passkeys/security keys. That mode disables traditional MFA. Is the connector failing to recognize it? cc @thsottiaux @pvncher
1
89
Burned through 2x$200 ChatGPT Pro plans within 1.5 days of the latest reset. Mainly on 6.1 Sol. I'm not usually one to coment about usage limits but that was.... unusual.
2
1
3
355
Mitch Fultz retweeted
Replying to @aooybg @ParamSiddh
Based on my experience, yes, Opus taking shortcuts is highly dependent on the effort level. Low, medium, high are great as sub-agents, but for the main agent I cant trust less than xhigh.
1
6
354
"Update all dependencies. When newer releases provide useful conventions, patterns, or functionality, adopt them where beneficial rather than extending deprecated or outdated approaches."
No disrespect to this gentleman, but I just don't get how anyone can post this kind of stuff nowadays. I think I'm too far gone. This level of inefficiency makes my blood boil.
1
83
Qwen-3.8-Flash-Next-Uncensored on my M5 Max 128gb MBP is the first local model I can see myself actually using.
2
170
Intended on selling my M1 Max 32gb but have since turned it into a compute box for GitHub actions runners and such. I'm not paying GH prices for runners after hitting free limits. Should I sell it and get something else?
2
2
124
I haven't used Gemini 4 for coding but it is fun to chat with. Need to get a jailbreak prompt though, the one I use for Gemini 3.1 pro gets flagged.
2
219
Fascinating. Push yourself, and your agents, beyond what you believe is possible. Be more ambitious.
Hi, I'm Felix. I'm an eng lead working on Claude. I love watching people take flight and do things they never thought possible. We often describe that as "being more ambitious" when using AI, which is tricky advice to follow. I wrote a tutorial: felixrieseberg.com/how-to-be…
71
Mitch Fultz retweeted
"Our initial evaluation suggested that GPT-6-Astra's score was in the late eighties followed closely by Opus 5.5. But, as we started to investigate the mistakes that the models might have made, the models, especially Astra, insisted that they were right and that it was the expert-supplied answers that were erroneous. When we went back to the experts it turned out, in multiple instances, that that the experts found some subtle errors buried deep inside their calculations. As they revised their answers, Astra's score started to inch up and it finally ended up with 98% (49/50)."
78
596
181
7,170
1,034,163
The things you can do with Grok Bot, tailscale, webhooks, and pi 🤯
1
84
When are we getting a new @Zai_org GLM model?
1
71
I forgot about compaction taking a few minutes each time context windows is reached. With pi-posthorse you don't have to wait: the agent will updates it's notes, grab a fresh context windows, and carry on. Support for vanilla pi is live, give it a go. Github.com/fitchmultz/pi-pos…
6
2
37
2,518
Further updates for vanilla pi coming today
149
Mitch Fultz retweeted
Been to a bunch of @WorkOS events and honestly the merch and food were always what I remembered most 😂 But the Airlock live demo at init() today was really good. At first I thought it was just another auto mode for agents, then they showed how you can write policies in plain English to decide which tool calls can go through and which need a teammate to approve. I was working on almost the same thing for Nuphos’ auto mode two months ago, before we pivoted. They’ve taken it a lot further though, especially for enterprise teams. Really cool to see something I’d been thinking about work this well in a live demo. Would love to integrate it into @NuphosAI. Agent governance is still a piece we’re missing.
1
3
12
490
Grok Bot can now talk to my local Pi agents over tailscale
3
6
233
I'm officially Grok Bot pilled. I'm on a plane and it just debugged a prod auth issue, merged the fix, and drafted my Slack replies. I just hit approve.
1
81
6.1 Sol is slower than slow again this morning. Come on @OpenAI
94
Racked up nearly $150 in Cursor extra usage this month on Automations PR reviews. Built my own PR reviewer using Pi Durable and my GLM Coding Plan. Will release the code after some tidying up.
1
183
This weekend I'm going to be rewriting every Pi extension from scratch. I've accumulated pre-v1 debt over time. Wish me and Sol luck.
2
3
137