Every idea, framework, and brainstorm I've put into writing — ordered by when I first thought it. 12351 items total. Click any title for the thesis card; the card links to the full essay if you want it.
Capital allocation is fundamentally a discipline of matching investment intensity to the underlying economics of a business, and the atoms collectively reveal that the most consequential decisions are often what to refuse, not what to fund.
Information flow is the central nervous system of an organization, and when it is restricted, hoarded, or filtered through hierarchical permission gates, nearly every operational failure — missed deadlines, misaligned priorities,
T3 concept. 4 atoms reference this.
T3 concept. 4 atoms reference this.
Prompt engineering is evolving from one-off questioning into a deliberate discipline of structured, reusable instruction design.
T3 concept. 4 atoms reference this.
Sales psychology is less about clever scripts than about reframing the buyer's experience of commitment, friction, and their own emotional signals.
T3 concept. 4 atoms reference this.
T3 concept. 4 atoms reference this.
good training philosophy is defined by deliberate, evidence-based defiance of inherited orthodoxy rather than reflexive adherence to it.
T3 concept. 4 atoms reference this.
Cumulative culture — the ratchet-like process by which human knowledge accumulates, transmits faithfully across generations, and compounds over time — is what genuinely distinguishes our species, not raw individual intelligence.
listening is a strategic superpower built on the deliberate redirection of attention away from yourself and toward the other person, and its power comes from the fact that almost nobody actually does it.
Nathan does not treat the AI's first draft as the deliverable.
Root-cause thinking requires resisting the pull toward visible symptoms and instead tracing problems back to their upstream origins, where a single underlying mechanism often generates many seemingly unrelated effects.
Three recurring failure shapes share the same structure: the user has authoritative knowledge the LLM lacks, and the model substitutes a defensive inference instead — 'it's premium,' 'we're building X not Y,' 'it was working before, you
For YouTube and most major media hosts, the oEmbed endpoint only returns structured metadata for genuinely public assets — paywalled or private videos simply don't respond.
When a subagent can't immediately find a public source, it hedges with labels like 'premium-only,' 'paywalled,' or 'no public source' — an authoritative-sounding scarcity claim that lets it avoid the harder admission of 'I couldn't find
When a user has reasoned their way to "upside exceeds downside" once, downstream extensions of that license don't require re-reasoning. "do both" (2 words) extends a license originally granted as "yeah why not" (3 words).
Peter Steinberger's monthly reminder reframes the entire job of working with coding agents: don't prompt them yourself, design the loops that prompt them.
Briefings compress detail into single numbers like '171 pending,' and that number cannot tell you whether the pipeline is healthy-and-busy or stuck-and-silently-failing — same value, opposite action.
For any metric reported in a briefing three or more days running, the persistence is the signal — not the absolute value.
Photon isn't the optimal iMessage bridge for a Mac user — but it IS the optimal 'phone-to-desktop-app-when-away-from-Mac' tool, because it's the channel that already reaches the Hermes gateway. Optimality is multi-axis.
When an agent or doc pushes back before executing, it's usually because the surface request ('set up Photon') is mismatched with the underlying job ('let my phone reach Hermes when I'm not at my Mac').
Set the readability test at the hardest realistic viewer: 'a parent reading on iPhone must understand the chart without zooming or squinting.' That single sentence fixes font sizes (≥11pt), legend placement (no collision with data), and
Numbers buried in narrative paragraphs do storytelling work, but they can never show pattern across repetitions. The clearest case: a swimmer's back-half fatigue trending from 37.1s to 38.5s across nine 50m reps.