X PAPERPaper mode
ArchiveLatest issue
Menu
—
Reading controls
Drag
Move the paper
Double-tap
Zoom in
Tap, then hold
Slide up to zoom in, down to zoom out
Triple-tap
Fit the whole page
Drag a paper corner
Pull far enough to turn; release early to return
Mouse and keyboard
Scroll to zoom, arrows to move, 0 to fit
Newspaper page
THE PERSONAL DAILY Tuesday, October 13, 2026 Issue 008 / Edited edition
From your following
Edited for reading

X Paper READ

Your following. Your daily paper.

Take time to read
See who said what
81 followed sources · This issue 14 stories Named voices · Traceable sources Latest issue
IN THIS ISSUE

OCT
13
Finance & Business

Economics prize recognises Hall and Woodford’s foundations for stabilisation policy

On October 12, the Royal Swedish Academy of Sciences awarded the 2026 economics prize to Stanford’s Robert E. Hall and Columbia’s Michael Woodford for the microeconomic foundations of macroeconomic stabilisation. They will share 12 million Swedish kronor equally.

The Academy describes the 1970s oil crises as a challenge to older policy models: unemployment and inflation rose together. The laureates rebuilt macroeconomic models around decisions made by households and firms.

Expectations are central. Households consider the future when saving; businesses do so when setting prices. Announced government and central-bank plans can influence those expectations, alongside the effects of current interest rates and spending.

The Academy credits this framework with improving stabilisation policy. That is its assessment of the research contribution, rather than a new central-bank decision or a claim that models can prevent every crisis.

edited

Finance & Business

IonQ agrees to buy Elisity in stock deal worth up to $1.3 billion

IonQ announced a definitive agreement on October 12 to buy Zero Trust security company Elisity. The all-stock transaction is valued at up to $1.3 billion, including up to $150 million tied to Elisity’s performance.

Closing is expected in the fourth quarter of 2026, subject to regulatory approval and other customary conditions. Shares will be calculated at closing using the agreement’s specified 20-trading-day volume-weighted price, bounded between $33.07 and $49.61.

Elisity uses identity, asset and risk information to restrict communications by users, devices, workloads and AI agents, aiming to contain intrusions. IonQ plans to combine these controls with post-quantum cryptography and quantum key distribution.

The companies expect cross-selling and integrated products; Elisity’s existing leadership is to remain after closing. Those benefits and security outcomes are plans, rather than independently verified results.

edited

AI与科技

EmbeddingGemma 2 brings multimodal retrieval to local devices

Google DeepMind’s model card clarifies the EmbeddingGemma 2 shared by Gorden Sun: it maps text, code, images, video and audio into a shared vector space for retrieval. It is an embedding model, rather than a chatbot.

The full model has 740 million parameters; the text encoder alone has 270 million. Its 768-dimensional output can be truncated to 128, 256 or 512 dimensions. The shared context limit is 8,192 tokens, including media.

The Apache 2.0 open-weight model targets local devices. It differs from the cloud-based Gemini Embedding 2 API; local use still requires suitable hardware and an indexing pipeline.

The card specifies bf16 or fp32, warning that fp16 can cause NaNs or degraded results. Queries and documents require different task prefixes. Truncated vectors need renormalization, and retrieval quality must be tested on the intended collection.

edited

财经与商业

Tencent proposes international notes; completion remains conditional

Tencent’s October 12 announcement supports the bond proposal highlighted by fxtrader. The company plans an international offering to professional investors, but no binding issuance agreement had been signed at the announcement date.

The proposed notes may have different series, maturities and currencies. Issuance depends on market conditions, corporate needs and investor interest. A programme limit is not the amount actually raised.

The announcement does not by itself establish the post’s roughly US$5 billion figure as a final issue size, or prove a dedicated AI spending allocation. Final pricing and use-of-proceeds documents remain outstanding in this review.

A separate pricing announcement on Tencent’s investor site is dated June 10. It cannot prove that the October proposal has completed. Proposal, pricing and closing are separate stages.

edited

AI与科技

Meta launches Muse Wonder, with different access routes

Meta’s October 12 announcement confirms the release highlighted by dotey. Muse Wonder is available through the Meta Model API and Muse Code; Muse Wonder Full is available to Muse Code subscribers. Access is not identical across versions.

Meta emphasises Active Memory, carrying relevant information across conversations, and Active Focus, pursuing goals through obstacles and checking before important actions. These are company descriptions, not independently reproduced results in this report.

The consumer Muse rollout is described as coming soon, first to subscribers. Developer availability does not establish that every consumer account already has the same model.

The announcement makes no open-weight commitment and says fuller safety evaluations will follow. Claims of frontier performance require task-specific and independent assessment rather than treating launch branding as universal superiority.

edited

Finance & Business

Pension spending: budgets and different populations do not compare directly

fxtrader challenged a circulating pension table. The Finance Ministry’s 2025 final accounts list administrative and public-institution pension spending at RMB1,733.476 billion; RMB1,604.7 billion was the budget.

Against final social-security and employment spending of RMB4,441.691 billion, the share is about 39.03%. Our table check covers 2025 only.

The category includes public-institution staff and contributions for current employees. Enterprise-worker and resident categories mainly record fiscal subsidies to insurance funds. These different accounting scopes cannot establish a per-person pension gap.

edited

Finance & Business

Trade in kilos: a useful lens with important omissions

In a three-post thread, Brad Setser says Europe now imports more machinery, electronics and cars by weight from China than from the rest of the world combined.

He says European exports of those goods to China weigh roughly half their level six years earlier, while corresponding Chinese exports are nearing twice their earlier weight. We have not independently recalculated his charts.

Weight separates some price effects but does not measure value, technological content or services. Setser acknowledges that internet advertising cannot be weighed. The charts describe goods flows, not total gains from trade or proven employment losses.

edited

Finance & Business

Sahm: explain the decision on decision day

Claudia Sahm’s October 12 essay criticizes the Fed’s communications: the public waits three weeks for minutes to understand the committee’s reasons more fully. She supports the earlier increase itself.

She distinguishes tightening because the economy is stronger from a small adjustment to insure against inflation risks. A common vote need not mean a common rationale; an abbreviated statement can leave markets misreading the policy path.

Her proposal is to state the main reasons and disagreements immediately, leaving detail to the minutes. This is an economist’s recommendation, not a new Fed procedure or a prediction of the next rate decision.

edited

AI & Technology

Self-improving AI needs a cost ledger, not just a score

a16z’s interview excerpts with MakerMaker.AI’s founders pose two tests for recursive self-improvement: who optimizes whom, and whether performance gains survive the cost calculation.

People using AI to improve AI are different from a model improving its own weights, harness or task prompt. The team says it tracks all experiment spending and assigns an engineer to challenge results and identify variance.

The founders report early signs of better and cheaper operation, but the excerpts supply no reproducible experiment package. We reviewed three published excerpts, not the whole video, and treat the unpublished results as their claims.

edited

AI & Technology

A larger Codex context: why fewer compactions matter

During this issue’s window, Baoyu shared Tibo’s announcement from the previous evening: GPT-6.1 Sol’s default context in Codex and Work rises from 272k to 500k, with a 1M maximum.

The default and ceiling are different. Tibo says testing supports the new default to reduce compactions and preserve detail in long tasks; users can override it. This does not mean every request uses 1M or subscription allowances double.

The original announcement predates our window by 15 minutes; the new activity here is Baoyu’s explanation. We did not test account settings, latency or task accuracy. Fewer compactions are the provider’s stated goal.

edited

AI & Technology

Python 3.15: useful changes without an automatic upgrade mandate

Simon Willison shared practical notes on Python 3.15. Official documentation dates its release to October 9 and describes explicit lazy imports and the profiling package.

Lazy imports defer loading until first use, potentially reducing command-line startup work; import failures may likewise appear at first use. The Tachyon sampling profiler observes running programs.

Willison also discusses using Python 3.11 features now that his own projects no longer target 3.10. That is a maintainer’s compatibility choice, not an instruction for every project. We did not benchmark performance or validate dependencies.

edited

AI & Technology

Claude Code usage: identify the plugin and the output first

Replying to a user, Boris Cherny suggests disabling the up-coder plugin to see whether usage improves; he says it appears to consume most of the tokens.

Nearby replies request the complete /usage output and identify another pasted output as not coming from Claude Code. Diagnosing usage needs recognizable logs and a clear scope; displayed numbers may belong to different tools.

This is advice for an individual case, not evidence of a universal defect, a completed fix or a quantified saving. We do not generalize the support exchange into a product-wide finding.

edited

AI & Technology

Tibo’s Pro discussion weighs allowance against pacing

Discussing a Pro-user poll, Tibo says he expected stronger support for twice the allowance even if five-hour limits returned to smooth demand.

He mentions Pro 100, 200 and 500. Stored evidence lacks the complete poll choices and vote counts, so we report no winning share. Total allowance and short-window availability are different constraints.

This is a discussion of user preferences, not an announced change to pricing, allowances or five-hour limits. The practical issue is uninterrupted work versus a larger total budget.

edited

Arts & Life

OneRepublic points fans to a full Hollywood Bowl clip

OneRepublic’s official account says its full Hollywood Bowl performance clip of “West Coast” is available in the band’s app.

The post does not date the performance or announce a new album or tour. We did not watch the full clip; the useful update is where fans can find the complete recording.

edited