Generate and share music about AI risk

Project summary

i make songs, poems, raps and music videos about ai risk, in spanish and english. the project is called feel the agi.

my thesis is that we need more people who make others feel what is at stake with ai, not only understand the arguments. a person can agree that losing control of powerful ai would be dangerous and still feel that it has nothing to do with their life. i want to bring that possibility closer through music, stories and a human performance.

i am seeking $5,000 for a 12-week pilot: six pieces, including higher-production versions of party in the ai lab and titubeo, with short videos to help people discover them. i make two versions: one with an ai voice, and another with my own voice and face. the work is bilingual across the project; not every song is promised as two separately sung language versions.

examples:

- [party in the ai lab](https://youtu.be/b8yeQnP_ftw)

- [titubeo short](https://youtube.com/shorts/4r5WOVhweJQ)

- [ia, démonos tiempo](https://youtu.be/423JhEoWBfM)

- [petrov and ai-risk talk](https://youtu.be/MnYnl9Bpcds)

What are this project's goals? How will you achieve them?

Reach 10K listeners. I will run social media campaigns.

## what are this project's goals? how will you achieve them?

the production goal is six songs, poems or raps with accompanying videos over 12 weeks. the human goal is for someone outside the ai-safety community to hear a piece, understand a concrete risk and remember it well enough to explain it to someone else.

my reach target is 10,000 listeners or viewers during the pilot. that is a target, not a forecast. i will report platform-estimated unique audience where available and views separately, rather than counting every play as a different person. i cannot reliably deduplicate people across platforms.

i will write about loss of control, the ai-development race and what happens to human agency as we delegate more decisions to machines. i will use music and recognizable situations as an entrance, without presenting speculation as established fact. some pieces are parodies of familiar songs; others are poems or original work.

each release will have short excerpts for discovery, alongside the full piece. i will distribute through my existing channels and relevant communities where sharing is welcome, and ask for specific feedback about what people understood. this budget does not include a paid advertising campaign.

i will review the work at weeks 4, 8 and 12: what shipped, who watched, where people stopped watching, and whether anyone accurately retold the idea. if the first releases do not connect, i will test different openings, shorter versions or a different format before spending more on polish. views alone will not establish that the project reduced ai risk.

How will this funding be used?

the main request is $5,000 for three months, without assuming another funding source will pay for the baseline costs:

| expense | usd |

| --- | ---: |

| claude max, $100/month for three months | 300 |

| baseline fal video credits, $100/month for three months | 300 |

| additional production for the two flagship videos | 3,500 |

| living allowance for three months | 440 |

| explicit contingency | 460 |

| total | 5,000 |

the additional production allowance is for video generation beyond the baseline credits, animation or editing help, and audio mixing. it is a planning estimate, not an agreed contractor quote. the living estimate is based on about s/500 per month and an approximate currency conversion. no equipment purchase is planned.

with about $1,000, i could cover the baseline tools and part of my living costs and make simpler versions, without the additional flagship production. the larger request is primarily a production budget, not mostly a personal stipend.

Who is on your team? What's your track record on similar projects?

it is currently just me, luis fernando yupanqui, an independent ai engineer and creator based in cusco, peru. i do the writing, recording, production and publishing, using ai tools as part of the process.

my youtube channel has about 1,630 subscribers. party in the ai lab has about 3,800 views; nine spanish shorts from titubeo have about 6,660 combined. i have also published ia, démonos tiempo and a talk about petrov and ai risk. most of my full-length songs still have small audiences. i can show that i finish and publish work; i have not yet shown that this music consistently reaches a large audience.

i also made a physics short that reached roughly a million views. that is evidence of broader video experience, not evidence that a million people watched my ai-risk work.

outside music, i built production computer-vision systems at deepview handling over four billion images per year. i ran so long sucker, a public ai-deception experiment with 605 human-vs-ai games and a hacker news front-page launch. my symbiotic ai project has about 700 github stars. this engineering experience helps me work with the tools and the subject matter, but it does not remove the uncertainty about audience response.

[website](https://yupanqui.xyz/) · [github and professional background](https://github.com/lout33) · [youtube](https://www.youtube.com/@yupanqui5)

What are the most likely causes and outcomes if this project fails?

the clearest failure would be finishing six pieces that people do not choose to watch, or that only reach people already concerned about ai. another is making something catchy that leaves listeners with a misleading impression rather than a better understanding. better production alone may not fix either problem.

there are also execution risks: recording and animation may take longer than planned, and parodies can receive copyright claims or distribution restrictions. the ai-voice and own-voice versions will be identified clearly, rather than implying that every existing release is a human vocal performance.

if the early pieces do not connect, i will use the remaining time to test cheaper versions and revise the approach rather than automatically spending the full production allowance. i will report what failed. if i cannot secure funding or a paid lead by january 4, 2027, my current fallback is to start paid engineering work and reduce the time i can spend on this project.