I meant to write after the lab dinner Wednesday before break, kept delaying, then figured I’d wait until after the paper submission—finally today, dorm packed, I have time.

Not sure what to write; felt busy but unclear on what was busy. Treat this as consolidating what I learned and sketching a plan.

Baseline: AI is a capability multiplier, so we should spend more time learning and stockpiling ideas. Main thread stays learning—deeper, more detailed. Internships/jobs stress me out, but part of me thinks they’re not that important—they’re external scorecards. Joining a company means shipping their product and their ideas. Why not try building your own—anything, good or bad. Walking in with a portfolio might help recruiting too.

Full stack (for now: front end)

Learning front end—TypeScript, React, small projects. Ideas not started yet; that’ll cost time, revisiting basics, and probably interview cram next term. End goal is full stack: web → app → backend/DB, step by step.

Agents

My agent knowledge still feels shallow—want to go deeper: Bojie’s courses/streams, LangChain-style frameworks, SDKs, APIs, hooks, skills, keep up with new tooling.

Including “frontend skills” and similar drops from seniors.

Memory

Including memory as a subtopic—it matters. I used to think “filesystem or RAG” and that was shallow. Internship + Bai Ting’s memory survey paper + many memory products showed there’s more to unpack.

Paper

Jan 29 ICML deadline—recently my senior and I cranked on the paper. I drew lots of figures/tables and got a lot of coaching on charts, tables, layout, prose—learned tons. All-nighter together—memorable. I need to internalize how he writes; next term more papers for graduation, but that can wait—move from “memory” to “disk” for now.

Internship

After the 8:00 deadline I printed internship paperwork; next day (the 30th, yesterday) I did a one-day trial at a memory startup—met cool people, learned a lot, saw products and their toolchain. For my own reasons I’m not going back next week—focus on my own stuff. Summer internship, day-to-day internship—I’m parking both and trying to build product first.

Internships feel like external validation loops—everyone says daily internships matter, summer ties to return offers—but the core is whether you can build and ship. OSS contributions too—pause for now; try your product. No time left—start creating. I want to apply with a portfolio. Fits the trend of AI shrinking the junior bench. Fundamentals still have to be solid.

Be open

I’ve been trying to be more open and talk to more people—working with my senior on the paper, even that one-day stint, reinforced it: interact more, assume less.

Trying different ways to connect. I want to shadow Kubernetes release team—prep starts now.

Also train expression—voice input (Typeless, Autotyper, etc.)—maybe it helps with stuttering.

Exercise

This morning I did squats—building doable habits matters. My girlfriend jokes “after 25 men are 65…”; my senior hammers exercise too—take it seriously.

In the end

That’s mostly it. Periodic summaries are nice—I type fast; lots of fluff.

Using Things for todos/project management—hope it sticks. Telegram for links to articles I’ve read.

Later I might organize notes on agent memory products.