We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Most compelling: a case study in complementary technical pairs
The Friendship That Made Google Huge — article · Author/creator: byline not surfaced; published by The New Yorker · Recommended by: Vinod Khosla. Khosla’s post points readers to the article and frames Jeff Dean and Sanjay Ghemawat’s side-by-side coding as a force that changed Google and the internet.
The article makes the recommendation useful rather than merely inspirational: Dean and Ghemawat took charge of making a fleet of unreliable machines behave as a resilient system, using checkpoints, encoding, compression, and low-level optimization; its explicit engineering lesson is that solving problems at scale requires knowing the smallest details. They later generalized that work into MapReduce, which let Google programmers use distributed machines as one computer and became the basis for Hadoop.
Key takeaway: find a technically compatible partner whose thinking is complementary—Dean says the right pair becomes a “complementary force.” Why it matters: this is a concrete case study in collaboration compounding into infrastructure, not just a founder anecdote.
The technical bottleneck: The Verifier Bottleneck
The Verifier Bottleneck — article · Author/creator: Vishal Misra · Recommended by: Martin Casado. Casado calls Misra’s post “another fantastic post” and says it uses recent math results to show the limitations of recursive self-improvement without new information. Misra’s post supplies the link to the full article and states its thesis directly: AI progress is limited by verification rather than computation.
The article’s central distinction is that recursive self-improvement can make proposals better and search cheaper, but only a verifier adds information about reality: “Compute buys proposals; verifiers buy knowledge.” In science, the verifier moves outside computation—a drug must survive a trial, a material must be synthesized, and an aircraft must fly—and AlphaFold illustrates how solving one search problem can expose verification as the next bottleneck.
Key takeaway: when assessing claims about self-improving AI, ask where the new information enters. Why it matters: Misra supplies a sharper test than model size or raw compute: distinguish cheaper recombination of existing knowledge from contact with an external verifier that expands what is known.
An infrastructure recommendation with a concrete political bargain
AI data centres are the future. Canada must overcome the backlash — opinion article · Author/creator: Kevin Yin · Recommended by: Tobi Lutke. Tobi shared the article and added his own blunt endorsement: Canada needs more data centres because they fit its technology and energy strengths, while calling the backlash “nonsensical.” The article identifies Yin as a Globe and Mail contributing columnist and Berkeley economics doctoral candidate.
Yin’s argument is more nuanced than a generic pro-build position. AI and its supporting hardware are becoming general-purpose inputs to science, business, defence, and health care; Canada therefore needs dependable computing infrastructure, data jurisdiction, and domestic capability to build and operate data centres. The article also acknowledges the local costs—higher energy prices, noise, relatively few permanent jobs, and difficult-to-tax profits—and notes that more than two-thirds of surveyed Canadians oppose data centres near their communities even while seeing a need for domestic AI infrastructure.
Key takeaway: treat compute as strategic infrastructure, but make the bargain visible: the article proposes favouring Canadian ownership or joint ventures, commitments to serve Canadian researchers and startups, and requirements that operators provide or finance equivalent power and grid capacity. Why it matters: it gives readers a concrete way to evaluate data-centre policy—who gets access to the compute, who pays for the power, and whether host communities actually share in the upside.
Ambition with feedback loops: Seeing Like a State
Seeing Like a State — book · Author/creator: not named in Tan’s post · Link: none supplied · Recommended by: Garry Tan. Tan uses the book to describe high modernism as faith in linear progress, technical mastery, rational production, and the ability to redesign nature and human nature for greater welfare; he also notes that it promised freedom from scarcity and natural calamity.
Tan accepts the book’s warning that ambition becomes catastrophic when “the map can no longer be corrected by the territory,” but argues that the warning was overlearned. His proposed synthesis is explicit: “Bring back high-modernist ambition, and add feedback loops,” with superintelligence-driven abundance potentially making those loops real.
Key takeaway: retain correction mechanisms without treating ambition itself as the error. Why it matters: this is the strongest conceptual reading lead in the set for thinking about AI deployment: build boldly, but keep reality able to revise the plan.