Lydia-2027

open-source your thought process

Approximate plan of my life right now:

November 2025 (CA):

Finish Inkhaven. Focus on posts about BCI/ML. Use Tinker for SvR research. Order sleep pod.

December 2025 (CA):

• Finalize SvR project → working paper

• Work out what’s happening in non-LM AI & interpretability thereof

• Keep reading https://curius.app/lydia-nottingham/read-later

• Research tamper-resistant safeguards for open-weight models

• Do some MathPhil prep & some travel. Visit Constellation

Returning to Oxford Mathematics & Philosophy

Why undergrad:

• I want freedom to research what I find most interesting & important

  • ← I want to do a PhD

  • ← Finishing undergrad will help

    • One eye open for PhD-equivalent-without-undergrad paths (noting what I care about is the autonomy/PIship)

      • Suggested in the kitchen to me today—hit up OP for 5-year grant? Low chance of success but should give serious shot

      • I know lots of ways to get funding for 1 year, but I care about the headspace enabled by a longer grant term

      • Also “what happens when I’m 30”

January-March 2026 (CA):

Resuming MathPhil studies

Frontload:

• Knowledge & Reality (epistemology + metaphysics)

• If I can, Philosophical Logic

  • If I can’t, some essay-based philosophy subject

• Finish LM research projects

The reasoning behind front-loading philosophy is that I’m not super excited about these courses, and if I get my least favorite stuff out of the way ASAP, the third (final) year of my program should be smooth sailing.

April-June 2026 (Oxford):

Sit exams in:

• Metric Spaces & Complex Analysis

• Linear Algebra

• 2/3 of: {Differential Equations I, Probability Theory, Measure Theory}

Take tutorials in Knowledge & Reality

July-September 2026 (CA):

• Independent research / ad hoc with institutions I think are awesome

  • + Write Philosophy of Mathematics essays

October-December 2026:

• Study Logic

• 2/3 of: {Functional Analysis I; Probability, Measure and Martingales; Information Theory}

• Take tutorials in Philosophy of Mathematics

January-March 2027:

• Study Set Theory

• 2/3 of: {Distribution Theory, Fourier Analysis, High-Dimensional Probability, Computational Complexity}

• Take tutorials in Philosophical Logic / other

April-June 2027:

Sit exams in:

• Logic, Set Theory, {4 more mathematics units}, Knowledge & Reality, Philosophy of Mathematics, {1 more philosophy subject}

Ideas to make Oxf tolerable (in the running):

• Formalize Lean for analytic philosophy

• Develop alter ego to sit philosophy exams

Ideas to make Oxf tolerable (discarded because I would love doing them & totally neglect my course):

• Oxford deeptech podcast

• Enthusiastic FFHII (Future of ‘Future of Humanity Institute’ Institute) continuation

Things to continue while at Oxf:

• Blogposting (cadence: ~once/month)

• Open-source life? Livestream / record 95% of day?

• ML/programming

To do:

• Map out 10-year, 20-year, 30-year, 40-year, 50-year 60-year plans. Check if satisfied with them → revise + iterate if not.

• Figure out 1-2 vectors I really want to push on.

  • Candidates:

    • Differential technological development

      • The theory

      • The practice

    • Safe superintelligence

    • Universal cognitive enhancement

    • Mapping ‘unknown unknowns’ FHI-style

      • AI for mapping ‘unknown unknowns’

    • Simulations

  • Positive template

Convo with ChatGPT

[Research A] + MathPhil + [Research B]. Can you remind me if there’s anything else for me to care about?

ChatGPT said:

Thought for 28s

you can only execute a handful of real projects at once before everything else collapses into maintenance noise.

[Research A] + mathphil + [Research B]. is already too many “main characters,” so the move is to merge and demote, not add:

1. collapse [Research A] + [Research B] into one thing.
   […]

2. relabel mathphil as a boss fight, not a life arc.
   it’s “clear the degree constraint with minimal flailing so future-lydia has optionality.” this is a finite project, not a narrative identity. it should not get a narrative slot; it gets a checklist.

3. add exactly one more protagonist: artifacts [..,']
   this is all your public-facing stuff: blog, talks, twitter, code, […] theorems, […] essays. not “content creation,” […]

everything else that actually matters sits under infrastructure, not “things to care about”:

• body + sleep = keep the hardware non-scuffed so 1 and 3 are possible.

• immigration / money / housing = keep the game board stable enough to not derail 1 and 3.

[…] different bucket: “[…]; no new long arcs.” that’s a hygiene task, not a project.

so if you want a literal answer to “anything else for me to care about?” in the narrative-bandwidth sense, my answer is: no. you get:

1. […] research […]

2. finish the degree constraint (mathphil as boss fight)

3. leave a visible trail of artifacts

everything else that feels important should be forced to justify itself as serving one of those three or it doesn’t get a storyline.