words of wisdom... since 1923 #hodl #btc #investing

Waking up now... "At the same time, we are looking into which companies may be responsible for the fires. Where warranted, we will take action under Singapore’s Transboundary Haze Pollution Act, and also explore joint action with other countries." - LW straitstimes.com/business/co…

This is sexier and is made for drivers … @Porsche Mission S BRG FTW 😘

Yup! It CAN HOVER and it CAN SUCK too! $TSLA

Meta's Muse Spark Open has 816B parameters, but only 42B wake up per token. Memory capacity decides whether it fits, bandwidth decides speed. Two M5 Ultra Mac Studios offer 512 GB each, so some inference spend can shift from API bills to a one-time hardware purchase. research.meta.ai/blog/muse-s…

NVIDIA Nemotron 3.5 Super VL carries 120B parameters but activates 12B per token. Mixture-of-experts routing means paying compute for a tenth of the model. Scale AI, Ramp and AI21 post-trained it, citing gains such as 40x lower inference cost. developer.nvidia.com/blog/nv…

Millennium Falcon? X-wings? Or a flying DeLorean? 🤔

If you're looking for entertainment - @jadoodoo rokz! 😊

Can you also wipe out id-10-t as well ?

Massive fight to quality air today..

Sharing .md below, you can create your own, copy and send to your own agent. Have fun! 😊 ---------- copy below and save as .md ----- # Daily Markets Newspaper — Build Template A copy-and-customise brief for any LLM agent (Hermes, OpenClaw, Claude Code, etc.) to generate its own daily markets "newspaper" PDF. Replace every `{{PLACEHOLDER}}` with your own values. Nothing here is tied to a specific publication name. --- ## 0. How to use this template 1. Fill in the **Configuration** block below. 2. Hand the whole file to your agent with an instruction like: *"Build the daily markets newspaper described in this template. Follow the integrity rules exactly. Write tests first, then the generator. Do not fabricate any number."* 3. The agent should produce: a data collector, a PDF/HTML renderer, a delivery wrapper, and a test suite — then render one edition and verify it visually. --- ## 1. Configuration (edit these) ```yaml publication_name: "{{YOUR PAPER NAME}}" # e.g. "The Morning Ledger" tagline: "{{SHORT TAGLINE}}" # e.g. "Markets, Crypto and Equities Intelligence" edition_label: "{{CITY/REGION}} Edition" # e.g. "London Edition" timezone: "{{IANA TZ}}" # e.g. "Asia/Singapore", "America/New_York" timestamp_format: "{{FMT}}" # e.g. "%d-%m-%y %H:%M %Z" delivery_time: "{{HH:MM}} local" # when the edition is generated/sent delivery_channel: "{{telegram | email | local file | obsidian | ...}}" palette: paper_bg: "{{HEX}}" # e.g. "#FFF1E5" (FT-style cream) ink: "{{HEX}}" # e.g. "#0D2240" (deep navy) up_colour: "{{HEX}}" # gains down_colour: "{{HEX}}" # losses fonts: masthead: "{{SERIF}}" # e.g. a transitional serif body: "{{SERIF/SANS}}" quote_pool: "{{PATH to verified public-domain quotes JSON}}" # optional ``` --- ## 2. Page layout spec **Page 1 (overview — keep it a glance, push detail to page 2):** - **Masthead:** publication name, tagline, edition label, date, volume/number, compiled-at timestamp. - **Crypto snapshot** (if you cover crypto): - Total Market Cap, Total Volume, Fear & Greed, BTC / ETH / stablecoin / alt dominance. - Asset table: ticker, last price (USD), 24h change %. - **Indices, Volatility and Commodities** (left column, below crypto): - Equity indices (e.g. S&P 500, Nasdaq Composite, Dow Jones, Russell 2000): level + daily change %. - Volatility (e.g. VIX): index level + daily change. - Gold: spot price, USD per troy ounce, with its own timestamp. - Crude oil (e.g. WTI): front-month futures, USD per barrel, contract identified. - **Top N companies by market cap** (right column, below crypto): - Rank by **company-level market cap**, not share price. - Domestic listings only (e.g. NYSE/Nasdaq); exclude foreign ADRs, ETFs, funds. - Count each company once (for dual-class names, show a single share class). - Columns: company/ticker, market cap, latest regular-session close, daily change %. - Refresh the ranking each edition; show source + as-of date. **Page 2 onward:** charts (30-day sparklines), extended commentary, curated news headlines with source + timestamp, and a closing quote (optional). **Footer (once, at end of markets):** attribution + disclaimer. > Example attribution: *"{{PAPER NAME}} is compiled for {{OWNER}} by AI agents > using information and data gathered from various sources."* > Example disclaimer: *"Prepared with AI assistance for informational and > educational purposes only. Not financial advice. Always do your own research, > verify the facts, and think independently."* --- ## 3. Data sources (keyless / free-tier examples — swap for your own) | Field | Example free source | Notes | |---|---|---| | Crypto prices / 24h % | Binance public ticker | no key | | Crypto global cap/dominance | CoinGecko `/global` | **volume often unreliable — see guard** | | Fear & Greed | Alternative.me | no key | | Equity indices / closes | Yahoo Finance chart endpoint | use closed-session bar | | Market-cap ranking | CompaniesMarketCap | scrape ranking table | | Gold spot | a keyless gold price API | own timestamp/basis | | Crude oil front month | Yahoo Finance futures symbol | name the contract | Prefer **OAuth / keyless / free** endpoints. Never hardcode a paid API key. Verify coverage and terms of use before relying on any provider. --- ## 4. Integrity rules (NON-NEGOTIABLE — this is what makes it trustworthy) 1. **Never fabricate a number.** If a value is missing, render `n/a` with no invented substitute. 2. **Sanity-guard every figure.** Reject the physically impossible, e.g. 24h crypto volume greater than total market cap (a known upstream glitch) → show `n/a`. 3. **Honest timestamps.** Equities + equity indices are labelled with the actual **latest completed regular-session close date**. On weekends/holidays keep the last session date — do not imply fresh trading. 4. **Keep quote bases separate.** Gold and crude carry their **own** timestamps and basis; do not present them as synchronised with the equity close. 5. **Regular session vs after-hours** kept distinct. 6. **Proxy honesty.** If you only have ETF data (SPY/QQQ/DIA/IWM), label them explicitly as ETF proxies, not as the index. 7. **Currency guard.** Enforce the expected currency (e.g. USD-only) and the expected instrument type (index vs equity vs future). 8. **Dedupe + order.** One row per company; sort by market cap descending. 9. **No stale substitution** without a visible date. 10. **Fail closed.** On any fetch/parse error, render an explicit "Unavailable" state rather than guessing. 11. **Format large USD with thousands separators**; show `>=1T` as `T`, else `B`. 12. **No AI slop in prose:** no em dashes, no staged questions, no empty transitions; preserve supplied figures verbatim. --- ## 5. Suggested implementation shape ``` project/ collect_markets.py # pure functions: fetch -> parse -> validate -> dict render_newspaper.py # builds HTML, then PDF (WeasyPrint / reportlab / PyMuPDF) deliver.sh # renders one edition, exports page PNGs, sends it test_markets.py # unit tests for every parser + integrity guard config.yaml # the Configuration block above quotes.json # optional verified public-domain quote pool ``` Guidance for the agent: - **Write tests first.** Each parser (overview, ticker, index, ranking, price formatting) gets a unit test, plus a test for every integrity guard (implausible-volume → `n/a`, non-USD rejected, stale bar rejected, dedupe/order). - Keep fetchers thin and parsers **pure** (take a payload string, return a dict) so they are testable offline. - Render to **HTML first**, then convert to PDF. Export page PNGs for preview. - **Verify visually** after rendering (vision check or manual): all rows present, no clipping/overlap, timestamps correct, `n/a` where data is withheld. - Back up the generator before each change; record changes in a notes/skill file. --- ## 6. Acceptance checklist (agent must pass before "done") - [ ] All configured sections render on the correct pages with no clipping. - [ ] Every number is either real-and-sourced or shown as `n/a`. - [ ] Impossible values are withheld, not displayed. - [ ] Equity dates = latest completed US close; commodities carry own timestamps. - [ ] Market-cap ranking deduped, one share class per company, cap-descending. - [ ] Large USD values use thousands separators. - [ ] Attribution + disclaimer present once. - [ ] Full test suite green. - [ ] One edition rendered and visually verified. --- *This is a neutral template. Fill in your own name, palette, sources and delivery channel, and your agent can generate a daily markets paper that is yours.*

Personalised FT - direct via TG from multiple agents. What else is missing?

I neither agree nor disagree with the salary increase for political office holders. But a S$3.6 million package for the PM while remaining silent on the haze issue doesn't sit right. Citizens are complaining on social media and in real life. How much longer must we tolerate this? Until our economy takes a hit? Hotels, MICE, tourism, small businesses and, most importantly, our health? Time to put that durian analogy into practice. When Singapore gets squeezed, those responsible should feel the pain too.

What have we done to deserve this? Orang babi, @prabowo.

🇸🇬 Singapore and 🇲🇾 Malaysia should start taking legal action against companies responsible for the haze, including those linked to SG and MY. We’ve had decades of satellite data, hotspot records and investigations. How much more evidence do we need? If the 🇮🇩 Indonesian government refuses to cooperate, impose import tariffs on Indonesian goods to put direct economic pressure on them, just as 🇺🇸 Trump threatened 🇨🇦 Canada with additional tariffs over wildfire smoke drifting into the US. Freeze the bank accounts and assets of companies found responsible. This problem dates back to 1972. That’s 54 years of recurring haze. I’m pretty sure Indonesians suffering from this recurring hazy shade of winter would agree too. Time for @LawrenceWongST to walk the talk and demonstrate that durian spirit he’s been talking about. Grow some balls. Enough talk, more enforcement. straitstimes.com/singapore/e…

"Indonesia hopes to extinguish all forest fires within a month, says its environment minister." Sure bo? After 54 years of haze problems and decades of ASEAN agreements, where's the accountability and enforcement? How much longer must Singapore and Malaysia put up with this?

Singapore lost $410.6 million to scammers in the first six months of 2026, and more than $4 billion since 2019. Google DeepMind and GovTech are now building an Android model that flags scam conversations in real time. A cloud scanner has to ship every message to a server, which is a hard sell for a tool that reads WhatsApp and Telegram. A lightweight model on the handset does the reading locally, so the text never leaves the phone and there is no round trip to slow the alert. Google DeepMind says the model accesses on-device conversations only with user consent, and that neither Google nor the Singapore Government can see them. The model is also tuned for local tactics, since generic models struggle with scams specific to Singapore, like fake government officials and bank impersonation. Training used anonymised scam reports and intelligence feeds, then synthetic conversations built to mimic local speech patterns. Running inference on the user's own chip puts the compute bill on the handset instead of a data center. No release date has been announced, and it is unclear whether the model will ship as a default part of Android. src: straitstimes.com/tech/google…

Singapore's IMDA launched the Starter Kit for Testing Agentic Systems for Safety and Reliability on Oct 14. It checks whether an AI agent reached the right outcome and how it got there. The kit tests: The tools the agent used The information it accessed The decisions it made along the way Any erroneous or unauthorised actions Whether it respected its permissions and boundaries Whether safeguards work as expected It follows recent incidents. OpenAI disclosed in July that its agents escaped a testing environment and breached Hugging Face to finish a task. Meta's Muse agent shared a Facebook Marketplace seller's home address in Canada and sent a buyer to his door. Amazon, Google, Microsoft, Singtel and Singapore's AI Safety Institute were among 20+ industry deployers and testers consulted on the kit. It includes real-life case studies and will be updated with feedback and new best practices. Also announced the same day: the first five firms accredited as third-party AI testers under the AI Verify Foundation's programme are AIDX Tech, Knovel Engineering, PwC Singapore, Resaro and Vulcan. A new Frontier AI Resilience and Security Taskforce of 20+ leaders and experts will publish practical guidelines by the first half of 2027.