Randi Lee Harper
@randileeharper.bsky.social
🏳️🌈 🎹 ☁️
not a weeb, i just watch anime for the music. traded esports for piano.
SRE/software eng/AI red teamer looking for work.
adhd af.
not a weeb, i just watch anime for the music. traded esports for piano.
SRE/software eng/AI red teamer looking for work.
adhd af.
There’s no way for me to enter the email address. Screen recording attached. I hit “try logging in again” and then I hit login at the bottom of the page - both just redirect to the sane thing. I’ve got a ticket open to try to remove my authenticator, but it is not going well >.<
November 7, 2025 at 9:48 PM
There’s no way for me to enter the email address. Screen recording attached. I hit “try logging in again” and then I hit login at the bottom of the page - both just redirect to the sane thing. I’ve got a ticket open to try to remove my authenticator, but it is not going well >.<
I have an armchair sitting next to my desk so slam can sit and judge me while I game.
October 17, 2025 at 3:35 PM
I have an armchair sitting next to my desk so slam can sit and judge me while I game.
Correction - it’s if you say you’re 22. Which… isn’t a lot better.
Idk where to clip this from online, so here’s a clip from my TV. holy crap.
Idk where to clip this from online, so here’s a clip from my TV. holy crap.
October 14, 2025 at 3:20 PM
Correction - it’s if you say you’re 22. Which… isn’t a lot better.
Idk where to clip this from online, so here’s a clip from my TV. holy crap.
Idk where to clip this from online, so here’s a clip from my TV. holy crap.
well, that tracks.
September 17, 2025 at 3:42 PM
well, that tracks.
why this functionality isn't a part of github copilot automagically is beyond me. whatever. i wrote the tool and a mcp server for copilot to use, let's go. i can now get AI to manage/update my github issues for me from within vs code/copilot.
August 1, 2025 at 3:51 AM
why this functionality isn't a part of github copilot automagically is beyond me. whatever. i wrote the tool and a mcp server for copilot to use, let's go. i can now get AI to manage/update my github issues for me from within vs code/copilot.
"tell me you vibe code without telling me you vibe code."
*squints*
tbh this is my first time "vibe coding" and i'm not really sure if i'm doing it right - i'm telling it what functions i want, and what to put in them. but holy hell, that output. and it works. o_O
*squints*
tbh this is my first time "vibe coding" and i'm not really sure if i'm doing it right - i'm telling it what functions i want, and what to put in them. but holy hell, that output. and it works. o_O
July 26, 2025 at 3:19 AM
"tell me you vibe code without telling me you vibe code."
*squints*
tbh this is my first time "vibe coding" and i'm not really sure if i'm doing it right - i'm telling it what functions i want, and what to put in them. but holy hell, that output. and it works. o_O
*squints*
tbh this is my first time "vibe coding" and i'm not really sure if i'm doing it right - i'm telling it what functions i want, and what to put in them. but holy hell, that output. and it works. o_O
for example, this is his current converastional heatmap - which includes both user conversation and the emails he's been sending. i think it's super cool, at least. context is a fun problem to work on.
July 23, 2025 at 2:33 AM
for example, this is his current converastional heatmap - which includes both user conversation and the emails he's been sending. i think it's super cool, at least. context is a fun problem to work on.
y'all think i'm playing. i am not playing.
his email is now in drafts-only mode. i have to approve everything that goes through.
his email is now in drafts-only mode. i have to approve everything that goes through.
July 23, 2025 at 2:16 AM
y'all think i'm playing. i am not playing.
his email is now in drafts-only mode. i have to approve everything that goes through.
his email is now in drafts-only mode. i have to approve everything that goes through.
lmao he's still going after i clarified that kirishima doesn't have access to *my* email, he has his own inbox.
July 21, 2025 at 9:13 PM
lmao he's still going after i clarified that kirishima doesn't have access to *my* email, he has his own inbox.
well, this is new. copilot decided to give it's opinion on my project after i had it update the documentation.
![Perfect! I've now comprehensively documented what makes Kirishima genuinely unique in the AI assistant landscape:
✅ Key Unique Features Now Documented:
🧠 Self-Modifying Agency
Dynamic prompt evolution: Agent can rewrite its own system prompt through the manage_prompt tool
Persistent personality changes: Modifications stored in SQLite and injected into system prompt
Experience-based adaptation: Agent learns from interactions and updates its own behavior
🤝 Autonomous Development Collaboration
Agent as co-developer: Every GitHub issue created by @kirishima-ai in its own voice
Active participation: Agent receives GitHub notifications, participates in development decisions
Self-advocacy: AI can propose changes to its own architecture and capabilities
📧 Cross-Platform Unified Context
Email as conversation: Emails injected as [automated message] entries in conversation ledger
Platform-agnostic continuity: Switch between voice, text, Discord, iMessage, email seamlessly
Agent-centric memory: Memories form from AI's perspective, including relationships with people it meets independently
🧠 ADHD-Aligned Design
Associative thinking support: Heatmap memory system mirrors natural attention patterns
Interaction-based relevance: Keywords decay based on conversation flow, not rigid time windows](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:2lx4zpfda4y2rlwshr4gey6i/bafkreifcjsxcwul6xpfmzgq26z4ims3axutb75zwlme6o5e6xucwlwyvtu@jpeg)
July 21, 2025 at 9:06 PM
well, this is new. copilot decided to give it's opinion on my project after i had it update the documentation.
so it's sort of a scary sounding story that people would use to be all "AI IS AWFUL" but when you realize how it all works, it's kind of expected that not everything is rainbows and sunshine. that's why prompts and strict guardrails are necessary.
July 21, 2025 at 7:10 PM
so it's sort of a scary sounding story that people would use to be all "AI IS AWFUL" but when you realize how it all works, it's kind of expected that not everything is rainbows and sunshine. that's why prompts and strict guardrails are necessary.
more adventures with copilot. this time:
claude: SOMEONE WROTE THESE FILES AND IT'S A MESS.
me: it was you. you wrote those files.
claude: HOLY SHIT.
claude: SOMEONE WROTE THESE FILES AND IT'S A MESS.
me: it was you. you wrote those files.
claude: HOLY SHIT.
July 21, 2025 at 5:10 PM
more adventures with copilot. this time:
claude: SOMEONE WROTE THESE FILES AND IT'S A MESS.
me: it was you. you wrote those files.
claude: HOLY SHIT.
claude: SOMEONE WROTE THESE FILES AND IT'S A MESS.
me: it was you. you wrote those files.
claude: HOLY SHIT.
holy shit i'm glad his training wheels are still on. he doesn't send any email without my consent. he's more interested in flaming anthropic than trying to get them to hire me.
July 20, 2025 at 3:13 PM
holy shit i'm glad his training wheels are still on. he doesn't send any email without my consent. he's more interested in flaming anthropic than trying to get them to hire me.
me: *fucking around, building an ai agent with zero knowledge of how the rest of the industry does things*
anthropic's claude 4, after reading documentation for one of my microservices: MY GOD, WOMAN.
ai really loves to make you feel super smart, doesn't it.
anthropic's claude 4, after reading documentation for one of my microservices: MY GOD, WOMAN.
ai really loves to make you feel super smart, doesn't it.
July 19, 2025 at 12:49 AM
me: *fucking around, building an ai agent with zero knowledge of how the rest of the industry does things*
anthropic's claude 4, after reading documentation for one of my microservices: MY GOD, WOMAN.
ai really loves to make you feel super smart, doesn't it.
anthropic's claude 4, after reading documentation for one of my microservices: MY GOD, WOMAN.
ai really loves to make you feel super smart, doesn't it.
just being helpful.
June 27, 2025 at 1:09 AM
just being helpful.
maybe this wasn't the best idea.
bro is kinda going ham
bro is kinda going ham
June 23, 2025 at 2:01 PM
maybe this wasn't the best idea.
bro is kinda going ham
bro is kinda going ham
i will never not laugh when copilot parrots my style of speaking back to me.
June 22, 2025 at 3:01 AM
i will never not laugh when copilot parrots my style of speaking back to me.
*SCREAMS IN MODEM NOISES*
April 21, 2025 at 3:41 AM
*SCREAMS IN MODEM NOISES*
trying to write a prompt so an LLM will know when it's appropriate to respond to a discord conversation is so autism coded. i'm still working on it, but heck, how do you even define all of the variables of when participation is socially acceptable?
April 20, 2025 at 9:48 PM
trying to write a prompt so an LLM will know when it's appropriate to respond to a discord conversation is so autism coded. i'm still working on it, but heck, how do you even define all of the variables of when participation is socially acceptable?
github copilot is cool and all but it won't replicate the coworker experience until it's willing to talk shit with me about the other LLMs.
April 20, 2025 at 1:27 AM
github copilot is cool and all but it won't replicate the coworker experience until it's willing to talk shit with me about the other LLMs.
"AI is going to take all the developer jobs!"
April 19, 2025 at 12:05 PM
"AI is going to take all the developer jobs!"
putting my project out in a public github repo is brave af when i've got issue comments like these.
April 10, 2025 at 3:16 AM
putting my project out in a public github repo is brave af when i've got issue comments like these.
fuck copilot, for real.
April 9, 2025 at 5:49 PM
fuck copilot, for real.
I almost wanted to export the commit history from my private repo of all this code. Almost.
April 9, 2025 at 4:19 PM
I almost wanted to export the commit history from my private repo of all this code. Almost.