18+

Nur für Erwachsene

DIRTY zeigt Inhalte für Erwachsene. Mit dem Betreten bestätigst du, dass du volljährig bist und solche Inhalte sehen möchtest.

Was DIRTY ist · Nutzungsbedingungen

Gallery Entdecken Feed Markt Nachrichten

AnmeldenBeitreten

DIRTY.socialWas DIRTY ist Entfernung & Meldung Nutzungsbedingungen Datenschutz legal@dirty.socialDas Netzwerk für Adult Creator · 18+

Entdecken
Feed
Nachrichten
Profil

Ruo Shui (@ruoshuiresearch) · DIRTY

Ruo Shui

@ruoshuiresearch

stay hydrated

215 Folgt27.180 Follower10.701 Posts

YouTube YouTube YouTube Bluesky

Ist das dein Profil?Übernimm es — dann gehört dir ruoshuiresearch.dirty.social.Profil übernehmen

Ruo Shui@ruoshuiresearch· 27d

A line graph titled "Figure 1. BRA Clearing Prices by Delivery Year for Major LDAs" displays capacity prices in dollars per Megawatt-Day ($/MW-Day) from the 2018/19 delivery year to the 2028/29 delivery year.

The vertical axis ranges from $0 to $350 in increments of $50, and the horizontal axis lists eleven delivery years. The graph tracks four regions (LDAs): RTO (orange dashed line), EMAAC (light green solid line), SWMAAC (dark blue solid line), and MAAC (light blue solid line with circular markers).

2018/19 to 2024/25: Prices for all regions fluctuate at lower levels, generally staying between $30 and $225. EMAAC generally has the highest prices during this period, while RTO dips to the lowest point (around $30) in 2024/25.

2025/26 onward: There is a sharp, synchronized spike in prices for all four regions. The lines converge and rise rapidly to approximately $270 in 2025/26, peaking near $330 in 2027/28, before slightly dipping to around $325 in 2028/29.

Source: PJM Capacity Auction results for 28-29

Ruo Shui@ruoshuiresearch· 28d

“Once, men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.”

“‘Thou shalt not make a machine in the likeness of a man’s mind,’” Paul quoted.

Ruo Shui@ruoshuiresearch· 28d

Data center hiring is concentrated in a handful of regions
Data center share of all Indeed job postings by core-based statistical area (CBSA), year to date June 2026. Labels mark the 10 CBSAs with the highest data center share.

Source: Indeed Hiring Lab. CBSAs with insufficient data are not shown.

A map of the United States shaded in different gradients of blue to represent the percentage share of data center job postings out of all local job postings.

A legend on the right side defines the shading categories for the Data center share of postings:
Lightest blue: 0–0.5%, Medium-light blue: 0.5–1%, Medium-dark blue: 1–2%, Dark blue: 2%+

Ten specific locations (CBSAs) across the United States are marked with a black dot and labeled with their respective percentages, indicating the highest concentrations of data center job openings:

Hermiston-Pendleton, OR 85%
Washington Court House, OH 55%
Rockingham, NC 39%
Bloomsburg-Berwick, PA 38%
El Campo, TX 18%
Columbus, OH 14%
Jackson, MS 13%
Reno, NV 12%
Washington, D.C. 8%
The Dalles, OR 7%

Growth in tech job postings is driven by AI

Source: Indeed. Annualized posted salary for postings disclosing compensation on annual or hourly terms.

A scatter plot analyzing the relationship between the growth of specific tech roles and their salaries.

X-axis (Horizontal): Median annual posted salary (measured in 2025 dollars), ranging from roughly $50,000 to over $200,000.

Y-axis (Vertical): Log-change in job postings, ranging from -0.5 to 1.5. A baseline line is drawn at 0.0, indicating no change in postings.

Data points are color-coded into two distinct categories:

1. Data Center Roles (Pink Dots)
Three major roles are highlighted in pink, showing significant growth despite varying salary levels:

Data Center Technician: Located on the upper left. It has a high log-change in postings (around 1.0) but a lower median salary (under $70,000).

Network Technician: Located near the Data Center Technician, with a log-change around 0.6 and a lower salary.

Data Center Engineer: Located toward the center-right, with a log-change of roughly 0.8 and a median salary of around $130,000.

2. Other Tech Roles (Blue Dots)
Numerous blue dots represent various traditional and AI-focused tech roles. Prominent labeled roles include:

AI Developer: The highest outlier on the chart, with a log-change in postings near 1.8 and a median salary around $140,000.

AI/ML Engineer: Located high on the right side, showing a log-change near 1.0 and a high median salary (around $150,000).

AI Architect: Positioned near the AI Developer with a positive log-change around 0.8.

Staff Software Engineer & Platform Engineer: High-salaried roles (approaching or exceeding $200,000) that maintain positive posting growth.

Lower Growth Roles: Several roles cluster below the 0.0 line, indicating a decline in postings. These include Technical Writer, Systems Administrator, IT Analyst, Database Administrator, Java Developer, and Back End Developer.

Ruo Shui@ruoshuiresearch· 28d

A financial line chart titled "VIXEQ vs. VIX – Single-Stock Vol Dwarfing Index Vol" displaying data "Since June 2014" through mid-2026.

The top section compares two lines over time: a light blue line representing "VIXEQ" (single-stock volatility) and a dark grey line representing "VIX" (index volatility). Both lines show significant peaks, notably a massive concurrent spike in early 2020. Towards the right side of the chart, representing 2025 into 2026, the VIXEQ line trends upward to a final value of 49.0, while the VIX line remains relatively flat, ending at 15.0.

The bottom section shows a shaded grey area chart tracking the "VIXEQ – VIX Spread." The spread fluctuates mostly between 0 and 15 from 2015 through 2023, begins rising steadily through 2024 and 2025, and sharply spikes to its highest historical level at "+33.9" in mid-2026. A small text box in the bottom right corner notes "VIXEQ – VIX Spread Percentile: 100%".

The footer notes the source as Bloomberg, compiled by Citadel Securities and GMI, as of July 12, 2026, stating figures are for illustrative purposes only.

A line graph titled "Implied Correlations at Historic Lows" with the subtitle "SPX 1M and 3M Implied Correlations, 20-Year Lookback," displaying data from 2006 to mid-2026.

The chart tracks two lines over time against a vertical axis ranging from 0 to 100: a light blue line for the 1-month (1M) implied correlation and a dark grey/black line for the 3-month (3M) implied correlation. The area underneath the lines is filled with a light grey shading. Both lines show significant volatility, reaching peaks above 80 near 2008, 2011, and 2020.

Since 2020, both metrics display a steady downward trend, dropping to their lowest points on the chart by 2026. Two horizontal dashed lines indicate these current historic lows at the bottom right: the blue line ends at 7.2 and the dark grey line ends at 3.4. A small box in the upper right corner notes "1M Percentile: 0%" and "3M Percentile: 0%".

The footer lists the source as Bloomberg, compiled by Citadel Securities and GMI, as of July 12, 2026, noting that figures are for illustrative purposes.

Ruo Shui@ruoshuiresearch· 28d

Ruo Shui@ruoshuiresearch· 28d

The little girl was still in Bluey pajamas.

No older than 3, she was trying to smell the flowers.

At her side, a pink rolling bag.

Related Quote Block:
"Live updates: Biddeford shooting victim was not ICE’s intended target, officials say"

Her father, Joan Sebastian Guerrero, had just been fatally shot and yanked from a car he’d been driving. Now, he lay handcuffed in the street, where he would remain for five hours.

“You took her dad, you took her dad!” a woman nearby yelled, according to Cecelia Humiston, 22, who witnessed the interaction in Biddeford on Monday morning.

It is the second time in a week that federal Immigration and Customs Enforcement agents have stopped the wrong vehicle while executing an arrest warrant and fatally shot someone.

Agents were not wearing body cameras in either shooting.

Ruo Shui@ruoshuiresearch· 28d

forget majority of internet users...this idea is still unknown to majority of bluesky users

A diagram illustrates the main services and data flows involved in providing Bluesky. The graphic is organized into three sections:

Left Side (Users and PDS):

Users interact with Bluesky via actions like "likes, follows, posts, etc."
Their data flows into different Personal Data Servers (PDS):
Bluesky PDS (Bluesky-hosted servers)
Alternative PDS (third-party servers)
Self-hosted PDS (user-maintained servers).
The data from these PDSs flows into streams of records, relayed through systems labeled as "Relay" and "Alternative relay."
Center (Feed Generators and Labelers):

Firehose data (streams of records) is ingested by Labelers and Feed generators.
Feed generators create custom feeds or alternative views, categorized as "Bluesky App View" or "Other social mode."
Some data returns to alternative firehose systems.
Right Side (Services and Data Consumption):

Data flows back into different PDS types (Bluesky PDS, Alternative PDS, Self-hosted PDS) where users access it through:
"Counts, likes, etc."
Threads of replies, read views, and alternative app views.
Arrows show how custom feeds and app views flow between PDSs and feed services.
Overall Description: The figure highlights how Bluesky handles decentralized social media data, connecting user servers (PDS), feed generators, relays, and alternative app views. Various arrows represent data flows like user actions, feed generation, and display processes. Icons are sourced from Flaticon.com.

Ruo Shui@ruoshuiresearch· 28d

An line graph showing the 50-day average absolute 1-day percentage change for the Philly SOX (Philadelphia Semiconductor Index) from July 1994 to July 2026.

Y-Axis: Represents the percentage change, ranging from 0 to 6.

X-Axis: Displays a timeline with ticks every two years, starting at Jul-94 and ending at Jul-26.

Major Peaks: Significant spikes in volatility are labeled on the chart:

June 2000: A high peak approaching a value of 5.

Financial Crisis Late 2008: A peak reaching above 4.

COVID Crash: A sharp peak around Jul-20 reaching approximately 4.5.

Current Data Point: The line spikes upward at the far right of the graph toward Jul-26, labeled with the text +/-3.36%.

Ruo Shui@ruoshuiresearch· 28d

Marvell stock chart

Ruo Shui@ruoshuiresearch· 29d

@roryjohnston.bsky.social context

This is a dual-pane bar chart titled "Strait of Hormuz Transits: Commodity Vessel Count by Transit Type", dated July 13, 2026. The chart visualizes the daily ship counts passing through the Strait of Hormuz from February through mid-July.

The chart is divided into two main sections:

Left Section: West-to-East (Outbound) transits.

Right Section: East-to-West (Inbound) transits.

Both sides of the chart show a near-identical pattern divided by key chronological events marked by vertical dashed lines:

Pre-March (Baseline): A high volume of daily ship counts, fluctuating heavily between 50 and 80+ ships per day (represented by solid grey bars).

Late February / Early March ("Operation Epic Fury Begins"): A massive, immediate drop-off in standard shipping traffic occurs when this line is crossed. The ship count plunges to fewer than 5–10 ships per day.

Mid-April ("Ceasefire Announced" & "US Blockade"): Ship counts remain severely depressed, hovering mostly under 10 daily transits.

May ("Project Freedom"): Traffic sees a very slight increase but remains largely subdued.

Mid-June ("MOU Signed"): A noticeable upward trend begins.

July: Traffic experiences a moderate resurgence, peaking at around 25 to 40 ships per day in early July before dipping slightly.

Chart Legend and Transit Breakdown
Post-March traffic is color-coded to differentiate between transit types:

IMO Route (Green): Standard international routes, consistently present in small numbers after March.

Iranian Route (Dark Blue/Black): Appears more prominently starting in late June and July.

Dark Transit / AIS Transponder Off (Light Blue): Represents a significant portion of the traffic from March onward, indicating ships traveling without broadcasting tracking data.

Omani Route (Tan/Gold): Appears as a notable contributor to the traffic spike observed in early July.

The data is credited to Commodity Context, Kpler, and Bloomberg.

Ruo Shui@ruoshuiresearch· Jul 12

Friends: Repeat After Me

Ruo Shui@ruoshuiresearch· Jul 12

lol

Note: We’ve extended this promotion through July 19, 2026 at 11:59:59 PM PT. The 50% increase to Claude Code weekly usage limits has also been extended through the same date.

Ruo Shui@ruoshuiresearch· Jul 12

A horizontal bar chart titled "Grace-Blackwell GPU count (GB200 / GB300)" with a data cutoff date of 1 July 2026. The chart tracks the number of NVIDIA Grace-Blackwell superchips across various tech companies and national initiatives, broken down into three categories: Deployed (dark blue), Installing (bright blue), and Announced (orange). The horizontal axis measures the GPU count ranging from 0 to 600k.

HUMAIN (Saudi): The largest overall total, approaching 600,000 GPUs, almost entirely in the "Announced" stage, with a tiny sliver marked as "Installing".

Stargate Abilene (OpenAI/Oracle/Crusoe): The second largest total, around 450,000 GPUs, with a notable portion (roughly 60,000) currently "Installing" and the rest "Announced".

South Korea (national): Roughly 260,000 GPUs, entirely in the "Announced" stage.

Nscale to Microsoft: Around 200,000 GPUs, entirely "Announced".

xAI Colossus 2: Roughly 110,000 GPUs, entirely in the "Installing" stage.

Argonne Solstice+Equinox (DOE): Around 110,000 GPUs, mostly "Announced" with a small amount "Installing".

Stargate Norway (Nscale/Aker): Roughly 100,000 GPUs, entirely "Announced".

Several entries show smaller capacity deployments or installations (mostly under 50,000 GPUs each):

Together / Hypertec: Primarily "Deployed".

ByteDance (Malaysia, offshore): Entirely "Installing".

Stargate UAE / G42: Entirely "Installing".

xAI Colossus 1: Entirely "Deployed".

AWS Project Ceiba: Entirely "Installing".

Other entries with very small or early-stage counts include UK Nebius/Nscale, Mistral Bruno (France), Sines (Microsoft/Nscale, PT), Taiwan Foxconn, CoreWeave (GB300), IT4LIA (Bologna), Microsoft Azure GB300, Microsoft Azure GB200, HIVE/BUZZ HPC (Bell/Cohere), Indosat (Indonesia), and SoftBank.

The data tracks NVIDIA Grace-Blackwell superchips (GB200 and GB300), noting that one NVL72 rack equals 72 GPUs.

Sources listed: NVIDIA, operator disclosures, SemiAnalysis, Data Center Dynamics, and EuroHPC.

Ruo Shui@ruoshuiresearch· Jul 12

Claude Code hasn't been helping itself lately

A scatter plot comparing AI models across two metrics: Overall pass-rate (y-axis, ranging from 50% to 90%) and Cost per task (mean $) (x-axis, ranging from $0.50 to $4.50).

A red dashed Pareto frontier curve outlines the most cost-effective models for a given pass-rate, shading the area underneath in a light red gradient.

Key Data Points on the Pareto Frontier (Red Dots):
GPT 5.4-mini (codex): Near the bottom-left with a pass rate of roughly 58% at around $0.25.

GPT 5.4 (codex, med): Achieves a roughly 73% pass rate at approximately $0.65.

GPT 5.5 (pi, med): Reaches nearly 83% pass rate at just under $1.00.

Opus 4.8 (pi, high): Hits an 85% pass rate at exactly $1.00.

GLM 5.2 (pi): Reaches an 87% pass rate at around $1.30.

Opus 4.8 (claude code, high): Shows an 87.5% pass rate at roughly $1.95.

Opus 4.8 (pi, xhigh): Peak performance on the frontier line, achieving a 90% pass rate at approximately $2.20.

Sub-optimal Models (Dark Grey Dots below the curve):
Haiku 4.5 (claude code): The lowest performing model shown, sitting at a 51% pass rate for about $0.35.

Various configurations of GPT 5.5 (such as pi, high, codex, med, pi, xhigh, and codex, xhigh) and Sonnet 4.6 / 5 cluster in the middle, offering pass rates between 75% and 83% for costs between $1.00 and $2.50.

Several Opus 4.8 variations (like claude code, xhigh, pi, max, and claude code, max) are positioned on the far right, maintaining high pass rates (81% to 89%) but at significantly higher costs ranging from $3.20 to $4.30.

Ruo Shui@ruoshuiresearch· Jul 11

openrouter's fusion has been liking opus 4.8

An image showing a list of the "Top models used by Fusion," ranked from 1 to 5 based on volume or usage share:

Claude Opus 4.8 by anthropic — 33.8B (85.9%)

MiMo-V2.5 by xiaomi — 1.3B (3.3%)

Gemini 3.5 Flash by google — 1.1B (2.8%)

Owl Alpha by openrouter — 582.2M (1.5%)

Claude Fable 5 by anthropic — 434.0M (1.1%)

Ruo Shui@ruoshuiresearch· Jul 10

lol

by Editorial Team

Ruo Shui@ruoshuiresearch· Jul 9

An image titled "GPT-5.6 Sol enables digital work" displays five screenshots arranged in two rows, each demonstrating a different digital task.

Top Row:

Helping medical assistants: A screenshot of a medical billing and patient management software interface with various data fields and a search pop-up window.

Analyzing scientific data: A screenshot of a Jupyter Notebook interface displaying a Python code snippet and a boxplot chart analyzing protocol data.

Creating financial models: A screenshot of a detailed spreadsheet application filled with financial data tables, transaction overviews, and debt schedules.

Bottom Row:

Ordering groceries: A screenshot of an online grocery store website showing search results for broccoli florets alongside a sidebar of store departments.

Finding an apartment: A screenshot of an apartments.com listing page titled "Outlook DTC," displaying pricing, floor plan details, and availability for studio apartments.

Ruo Shui@ruoshuiresearch· Jul 9

Stacked bar chart from the IEA titled "Data centres account for much of the growth in US power demand," showing estimated electricity demand change by sector and region in TWh across three periods (2015-20, 2020-25, 2025-30). US 2025-30: data centres dominate at ~200 TWh, total ~400 TWh. EU 2025-30: transport electrification leads at ~150 TWh, data centres relatively small. Six categories: Data centres (gold), Industry (dark blue), Cooling (red), Heat pumps (pink), Transport (teal), Other (grey). Source: International Energy Agency

Ruo Shui@ruoshuiresearch· Jul 9

Ruo Shui@ruoshuiresearch· Jul 8

librarian would probably be top 5

An horizontal stacked bar chart titled "How Americans View The Morality Of Professions," subtitled "How would you rate the honesty & ethical standards of people in the following fields? [%]". The data is sourced from Gallup and visualized by Chartr.

Net Positive Professions
The top section features 11 professions viewed favorably on balance, ordered generally from highest to lowest positive perception:

Nurses: 76% High/Very high | 4% Low/Very low

Grade-school teachers: 61% High/Very high | 11% Low/Very low

Military officers: 59% High/Very high | 9% Low/Very low

Pharmacists: 57% High/Very high | 9% Low/Very low

Medical doctors: 53% High/Very high | 14% Low/Very low

Day care providers: 42% High/Very high | 6% Low/Very low

Funeral directors: 37% High/Very high | 9% Low/Very low

Police officers: 44% High/Very high | 20% Low/Very low

Auto mechanics: 33% High/Very high | 12% Low/Very low

Clergy: 30% High/Very high | 20% Low/Very low

Judges: 28% High/Very high | 26% Low/Very low

Net Negative Professions
A dashed line separates the top section from 12 professions viewed unfavorably on balance:

Bankers: 23% High/Very high | 31% Low/Very low

Local officeholders: 20% High/Very high | 32% Low/Very low

Nursing home operators: 21% High/Very high | 34% Low/Very low

Lawyers: 17% High/Very high | 33% Low/Very low

Newspaper reporters: 17% High/Very high | 45% Low/Very low

State officeholders: 14% High/Very high | 42% Low/Very low

Business executives: 12% High/Very high | 43% Low/Very low

Car salespeople: 7% High/Very high | 47% Low/Very low

Advertising practitioners: 8% High/Very high | 50% Low/Very low

TV reporters: 13% High/Very high | 55% Low/Very low

Members of Congress: 8% High/Very high | 68% Low/Very low

Lobbyists: 4% High/Very high | 68% Low/Very low

Ruo Shui@ruoshuiresearch· Jul 8

fun way to start an article

Over dinner at Mar-a-Lago, Trudeau’s envoys tried to dissuade Trump from absorbing their country. When one Trump aide pointed out Canada’s 41 million people would lean Democrat, the president came up with a neat solution: just split the northern neighbor into two states, one red, the other blue.

Ruo Shui@ruoshuiresearch· Jul 8

Issuance of Amended Iran-related General License

Release Date
07/07/2026

The Department of the Treasury's Office of Foreign Assets Control (OFAC) is revoking Iran-related General License X and issuing Iran-related General License X1, for the "Revocation and Wind Down of June 21, 2026 Authorization for the Production, Delivery and Sale of Crude Oil, Petrochemical Products, and Petroleum Products of Iranian Origin." Effective July 7, 2026, General License X, dated June 21, 2026, is replaced and superseded in its entirety by this General License X1.

Ruo Shui@ruoshuiresearch· Jul 7

that's some funky data 🤨

A line graph titled "Exhibit 4: In June, after-tax wage growth among lower-income households rose to 4.1% YoY, surpassing the 3.4% YoY growth rate for middle-income households." The chart plots after-tax wage and salary growth by household income terciles (Lower, Middle, Higher) from January 2023 through mid-2026, using Bank of America internal consumer deposit data.

Lower-income tier (dark blue line): Starts high in early 2023 at over 6%, steadily declines down to nearly 0.5% in early 2026, and then sharply rebounds to 4.1% by June 2026.

Middle-income tier (red line): Starts around 4.5% in early 2023, fluctuates downward to a low of nearly 1% in early 2026, and then recovers to 3.4% by June 2026.

Higher-income tier (orange line): Starts below 2% in early 2023, briefly dips below 0%, gradually trends upward to peak near 6% in early 2026, and then turns downward toward 4% by mid-2026.

Ruo Shui@ruoshuiresearch· Jul 6

A dual-directional bar chart titled "Temperatures Reach New Highs While New Lows Become Rare" with the subtitle "Ratio of record heat and cold compared to a stationary climate." The chart tracks data from 1960 to 2020.

A baseline horizontal line at the value of 1 represents a stationary climate (what would be expected without climate change).

Record Heat (Red Bars): Rising upwards from the baseline, the red bars show a massive upward trend over time. By 2020, the red bars peak at a value of 4. Text on the chart notes: "Record heat is now four times more frequent than expected without climate change."

Cold Records (Blue Bars): Extending downwards from the baseline, the blue bars show a steady decline over time, approaching the 0.5 mark near 2020. Text on the chart notes: "Cold records are half as common."

The source is listed at the bottom as: "Fischer, Erich M. (2025). Record-breaking extremes in a warming climate, Nature."

Ruo Shui@ruoshuiresearch· Jul 2

where next?

Oracle stock chart

Ruo Shui@ruoshuiresearch· Jul 2

An economic line graph from FRED (Federal Reserve Bank of St. Louis) showing data from 2005 to mid-2026.

Title: All Employees, Information / All Employees, Total Nonfarm

Y-Axis: Labelled "Thous. of Persons/Thous. of Persons", ranging from 0.017 to 0.024.

X-Axis: Timeline ranging from 2006 to 2026.

Source: U.S. Bureau of Labor Statistics via FRED®.

Trend Description:
The blue line shows a steady, long-term decline from approximately 0.023 in 2005 down to around 0.019 by early 2020. A grey vertical shading indicates the 2008 recession.

In early 2020, coinciding with a brief grey vertical shading for the COVID-19 recession, there is a sharp spike up to just over 0.020, followed by an immediate drop. The line then rises steadily through 2021 and 2022, peaking near 0.020 again. From late 2022 through mid-2026, the line shows a steep, continuous downward trend, ending at its lowest point of approximately 0.0175.

Ruo Shui@ruoshuiresearch· Jul 1

Bridgewater tinkered Qwen3-235B

A scatter plot titled "Figure 3: Accuracy versus price for our trained model and frontier models." The vertical axis represents "Average accuracy" ranging from 72% to 86%, and the horizontal axis represents "Cost per 1,000 tasks (USD)" ranging from $0 to $100.

A single green data point labeled "Our trained model" is positioned in the top-left corner, showing the highest accuracy at approximately 84.6% and the lowest cost at around $5.

Six grey data points representing "Frontier models" are clustered at lower accuracies and higher costs:

GPT 5.2: ~74.3% accuracy, ~$20 cost

Gemini 3.1 Pro: ~74.3% accuracy, ~$32 cost

GPT 5.4: ~75.8% accuracy, ~$29 cost

Claude Opus 4.6: ~77.2% accuracy, ~$61 cost

GPT 5.5: ~78.2% accuracy, ~$65 cost

Claude Opus 4.8: ~78.0% accuracy, ~$92 cost

The caption at the bottom reads: "Our model outperforms frontier models on both dimensions across generations."

Ruo Shui@ruoshuiresearch· Jul 1

Georgie crossing the Delaware

Ruo Shui@ruoshuiresearch· Jul 1

An stacked bar chart titled "Percent of Closed-End, First-Lien Mortgages Outstanding by Interest Rate," showing data from 2013 through early 2026.

The y-axis represents the "Percent of Outstanding First Lien Mortgages" ranging from 0% to 100%. The x-axis represents years from 2013 to 2027.

The chart categorizes interest rates into five color-coded brackets:

Dark Blue (<3%): Stays very low (under 5%) from 2013 until late 2020, where it surges dramatically, peaking at nearly 25% in early 2022 before gradually declining slightly through 2026.

Light Blue (3% to <4%): Hovers between roughly 20% and 35% from 2013 to 2020, then spikes alongside the dark blue section to peak near 40% in 2022, after which it steadily shrinks.

Grey (4% to <5%): Forms a large portion of the middle of the bars, remaining relatively stable between 20% to 40% throughout the entire timeline.

Yellow (5% to <6%): Starts as a thick band in 2013 (around 25%), narrows significantly to its thinnest point around 2022, and begins widening again into 2026.

Red (>6%): Starts quite thick in 2013 (over 30%), steadily narrows over the decade to its thinnest point in 2022 (less than 10%), and then expands substantially from late 2022 through early 2026 as higher interest rates return.

Ruo Shui@ruoshuiresearch· Jul 1

the shale ramp is something else

An line graph from the U.S. Energy Information Administration (EIA) titled "The 250-year history of U.S. energy consumption," plotting U.S. energy consumption by source from 1776 to 2025 in quadrillion British thermal units (Btu).

The graph tracks five energy sources, showing their historical trajectories and their 2025 percentages of the total energy mix:

Petroleum (37% in 2025): Depicted as a dark brown line. It begins rising sharply after 1900, experiences a notable callout stating "Petroleum surpasses coal in 1950," peaks near 40 quadrillion Btu around 2000–2005, and finishes as the highest source in 2025.

Natural Gas (36% in 2025): Depicted as a reddish-pink line. It rises steadily post-1950, briefly dips, and then climbs steeply after 2000 to finish just below petroleum.

Renewables (9% in 2025): Depicted as a bright blue line. It starts as the primary energy source in 1776, stays low and relatively flat for over a century, and has a callout noting "Coal surpasses renewables in 1885." It begins a steady upward trend after 2000.

Coal (9% in 2025): Depicted as a lighter pinkish-brown line. It rises rapidly after 1850, peaks around 2000–2005 at over 20 quadrillion Btu, and then declines sharply down to 9% by 2025.

Nuclear (9% in 2025): Depicted as a light blue line. It emerges around 1970, rises steadily, and flattens out around 8 quadrillion Btu heading into 2025.

The bottom of the image credits the data source to the U.S. Energy Information Administration's Monthly Energy Review.

Medien

Statistiken

27.2K

Follower

215

Folgt

10.7K

Beiträge

⭐ Top-Creator

Das bin ich — entfernen