“…to divide each of the difficulties under examination into as many parts as possible.” René Descartes |
|
| |
|
| 1 | | Lung-cancer combination cuts risk of death by 58% In a phase 3 trial of 563 patients with extensive-stage small-cell lung cancer, tarlatamab plus durvalumab extended median survival to 21.1 months, from 12.7 on durvalumab alone. One-year survival was 79.4% against 53.3%. WHY Small-cell tumors tend to return quickly after first treatment, so the trial used the combination as maintenance to hold them off. Severe side effects ran 46.8% against 40.6%; the step to watch is whether regulators extend the drug to this first-line setting. |
|
| 2 | | Retatrutide eases sleep apnea in two phase 3 trials Two 80-week trials of Lilly’s drug enrolled 691 adults with moderate-to-severe sleep apnea. In the trial without diabetes, the 9 mg dose cut breathing events by 34 an hour, against 10 on placebo. WHY Excess weight narrows the upper airway, so a drug that cuts weight can ease the obstruction without a machine at night. The step to watch is regulatory review. |
|
| |
|
S&P 7,580.07 −1.82% | NDX 28,384.9 −3.46% | 10Y 4.93% −6bp | BRENT $113.43 +2.37% | BTC $78.1k +0.04% |
|
| Intel’s weak forecast and Amazon’s negative free cash flow revived doubts about AI spending, giving the S&P and Nasdaq their worst day since June. Brent (January contract) rose on fears of escalation in the Iran war after tomorrow’s US midterms, then fell to $108.78 this morning as Iranian negotiators held a third day of talks with Qatari mediators in Doha. |
| |
|
| Two people from hard fields land in the same place. Larry Guth gets unstuck by asking a more basic question, not by waiting for the right one. Michael Nielsen spent two decades asking what he was, then found the answer in what he kept doing. In both, direction comes out of the work rather than before it. Esther adds the social version: closeness comes from what you keep showing up for, not from how convenient it is. |
Today’s question If you can’t think of what to build, are you short of ideas or short of a self? |
ESSAY arXiv · Larry Guth Guth, who works on Fourier analysis and the Kakeya problem, says doing math is mostly being confused and wrong, and that progress comes from asking something more basic. His example: the restriction problem stayed stuck until it was linked to a simpler question about overlapping tubes. Years of this left him crossing out any sentence he doesn’t really believe. On choosing work, he says not to stall waiting for the perfect choice, which feeds your note on running short of ideas that matter. |
ESSAY michaelnotebook.com · Michael Nielsen Nielsen argues creative identity comes from the activities you pursue, not a label like “writer.” He left academic physics in 2007 and was unsure what he was for two decades, until he saw writing as a tool for an explorer of under-examined questions. Darwin and Stewart Brand are his models: close observation first, an unusual form after. The risk he keeps open is dilettantism, and his test is new insight useful to others. This speaks to your note on shaping identity around what you value. |
ESSAY Curious Enough · Esther Esther argues friendship runs on mutual obligation, and that we now treat obligation as a harm. As convenience lowers tolerance for effort, unanswered messages get reframed as circumstance and the more accommodating friend quietly absorbs the cost. She leaves open what happens when someone really can’t meet the expectation. This pushes your question about convenience killing connection. |
| |
|
WORKFLOW karpathy.bearblog.dev Karpathy vibe-coded a pipeline in about 3 hours that gives an LLM ten years of hindsight on every front-page HN thread from December 2015. The shape is worth copying: download the data, bundle it into a markdown prompt, request grades in a fixed parseable format, render static HTML; 930 queries cost $58. The fixed format is the trick, because it turns one-off judgments into a score per commenter. His warning is the hook: cheap AI makes past public behavior easy to reconstruct. |
| |
|
AI + BIO anthropic.com Anthropic says six partners used Claude across discovery, manufacturing and trials: Enveda’s review caught all 11 issues humans found plus seven more, and one-person Mithril coordinated 11 outside labs. Nearly all of it is preclinical and self-reported. The claim to test is that bottlenecks are spread across the whole process, not just molecule design. |
MARKETS thediff.co The Diff argues robots will mostly take work that barely exists today, because it is too remote or unpleasant to pay people market wages for. Desert solar farms are the example: cleaning robots pay for themselves through recovered energy and fit existing project financing. The open question is whether displaced workers get a bridge to the new hubs. |
| |
|
Blog A Stanford math student arguing that as AI speeds up proofs, human understanding becomes the scarce resource. | | Newsletter Allie Volpe on friendship; start with the case for taking on small inconveniences on purpose. |
|
| |
Until tomorrow. Mon 02 Nov · 09:36 ET |
