“We perceive the world before we react to it, and we react not to what we perceive, but always to what we infer.” Frank H. Knight |
|
| |
|
| 1 | | Israel votes in its first general election since the October 7 attack Israelis vote today for the 120-seat Knesset in a race AP describes as a referendum on Prime Minister Benjamin Netanyahu, whose Likud faces a bloc led by former army chief Gadi Eisenkot. WHY Netanyahu presents himself as a survivor and coalition builder with religious and nationalist partners; the opposition says he must answer for the October 7 security failures and his refusal of a state inquiry. With 61 seats needed and no party ever reaching that alone, the next step is coalition talks that could run for weeks, with Mansour Abbas’s Ra’am a possible kingmaker. |
|
| 2 | | Regional envoys press Tehran for a deal as Egypt backs Saudi Arabia against the Houthis Saudi Arabia’s defense minister, Pakistan’s army chief Asim Munir and Turkish intelligence chief Ibrahim Kalin met Iranian officials in Tehran on Monday to push for a US–Iran agreement, while Egypt’s parliament approved military support for Saudi Arabia after escalating Houthi attacks. President Trump said it is “decision time” between a deal and renewed strikes. WHY A June deal for Iran to reopen the Strait of Hormuz in exchange for an end to the US blockade of Iranian ports unraveled; Iran says the blockade and sanctions must go first and rejects what it calls US demands for “complete surrender.” Trump has suggested strikes could resume after the Nov. 3 midterms; watch whether Qatari and Pakistani mediators settle the sequencing. |
|
| 3 | | FDA approves a targeted drug pairing for prostate cancer with DNA-repair mutations The FDA on Monday approved Pfizer’s talazoparib with enzalutamide for metastatic hormone-sensitive prostate cancer with HRR gene mutations. In the 599-patient TALAPRO-3 trial, the combination cut the risk of radiographic progression or death by 52% versus enzalutamide alone (hazard ratio 0.48). WHY Talazoparib is a PARP inhibitor: tumors that already cannot repair DNA through the HRR pathway lose their backup repair route and die, while hormone therapy attacks the cancer’s growth signal. Overall survival data are still immature, so the next readout is whether the delay in progression becomes longer lives. |
|
| 4 | | A painted-on liquid stopped tooth decay without drilling in a large child trial A Phase III trial of 830 children under six, published in JAMA Pediatrics, found that silver diamine fluoride applied every six months stopped decay in more than half of affected baby teeth, with no drilling, injections or anesthesia. WHY Silver kills cavity-causing bacteria and fluoride hardens the remaining mineral, which matters because cavities are the most common chronic disease in children and the hardest cases often end in surgery under general anesthesia. The trial supports an FDA application to approve the liquid specifically for arresting decay, which would move it off-label. |
|
| |
|
S&P 7,676.8 +0.70% | NDX — — | 10Y 4.98% −2bp | BRENT $115.81 −0.78% | BTC $79.8k −0.60% |
|
| Anthropic’s IPO filing revived the AI trade after weeks of losses, with tech and chipmakers leading the S&P 500 higher. Brent eased on hopes for Middle East diplomacy while attacks on Saudi Arabia limited the drop, and Treasuries held steady ahead of Wednesday’s Fed decision. |
| |
|
Today’s question When you size up a stranger in one glance, how much of what you see is them, and how much is you? |
ESSAY Letters from the Deep End · Jess A longtime true-crime follower realizes she had been passing confident verdicts on strangers built from fragments of news and speculation. The turn: after her first husband died, strangers judged her grief from the outside. Her sharpest line is that she had consumed so much information she confused it with knowing something. It pushes your note on misjudging people from incomplete information. |
ESSAY boundary · sublime The claim: close relationships are built by surviving differences, not by avoiding them, and friendships that never have a hard conversation quietly fade. The writer draws on Chinese “dazi” activity-friends and the Filipino cycle of tampo (sulking) and suyo (coaxing), plus Gottman’s finding that 69% of couple conflicts never resolve and lasting couples simply handle them better. It feeds your note that your purpose is relationships where friends feel safe to open up. |
ESSAY Terence Tao’s blog · Jeremy Avigad A logician and philosopher argues that AI can now produce results that once counted as respectable papers by recombining known techniques, so mathematicians must move toward harder problems, bigger self-chosen questions and new ways of using the tools. His mechanism: reinforcement learning rewards only final outcomes, which can yield cleverness without strategy, and no reward signal could have picked Riemann’s world-changing lecture topic. The open question he leaves is less whether AI can make Grothendieck-scale leaps than what humans value in doing math at all. It extends your note on building as an act of agency. |
Go deeper |
| |
|
TOOL hamel.dev · Hamel Husain Hamel tests Claude Code’s new build_eval and hill-climb commands, which propose likely failures in your app, have you label examples, build a grader, then iterate against it. The lesson that transfers to any agent you build: read your actual outputs before deciding what to test, because the failures that matter only show up in the data. His caveats double as a checklist for judging any “automatic” AI workflow. |
A first-impressions ledger Turn your note about being misjudged on the incline into a small tool that logs snap judgments and makes you revisit them. | 1 | Make a tiny form (a single HTML page or a Claude-built app): who, where, your first-glance read in one line, and a confidence score. | | 2 | Store entries in a local JSON file or SQLite, and have a scheduled script surface each entry again a week later with one prompt: what would the next frame have shown? | | 3 | After 15 entries, ask Claude to sort your reads by pattern and see which kind of judgment you revise most; that pattern is a video essay. |
|
|
| |
|
STARTUPS axios.com Anthropic filed its prospectus Monday, and the numbers show both concentration and depth: about 2,000 customers pay $1 million or more a year and 250 pay $10 million or more, yet the five largest made up about 25% of revenue in the first nine months. Nearly half of first-half revenue flowed through Amazon, Google and Microsoft cloud marketplaces, companies that are also its investors, suppliers and competitors. The open question is the almost $650 billion in computing commitments it must pay whether or not demand shows up. |
PREDICTION MARKETS theblock.co The NBA became the third league, after the NHL and MLB, to sign an information-sharing agreement with the CFTC, and named Kalshi and Polymarket its preferred prediction-market exchanges. It lands while states call the contracts unlicensed gambling and the NFL lobbies for state oversight. Leagues are choosing to legitimize the markets and police insiders from within rather than fight them. |
| |
|
Blog Short essays on doing math as AI gets good at it; this letter from someone who left the field recasts math as personal “mind-building.” | | Newsletter A daily digest of what AI builders are actually discussing; today’s issue rounds up reactions to Anthropic’s S-1. |
|
| |
Until tomorrow. Tue 27 Oct · 09:36 ET |
