"Richard Symons, the owner of a U.K.-based used-car sales company that specializes in EVs, has found that the batteries that power these cars (Teslas) continue to perform well even after several hundred thousand miles.
“They are proving themselves to be exceptionally reliable."
this guy just dropped a library of 25 INSANE loops for Fable 5
> automations for marketing, sales, research, coding…
> every one documented, flow mapped, prompt ready
> each paired with the exact tool to run it and the goal it hits
plus it's free... all you do is send these screenshots to Claude:
A BERLIN STUDENT PLUGGED CLAUDE + OBSIDIAN TOGETHER ON THE U-BAHN AT 10:14 PM. BY THE NEXT STOP SHE HAD A SECOND BRAIN RUNNING WITH THREE YEARS OF NOTES INSIDE
3,247 notes. 4 setup steps. 0 lines of code
downloads Claude Desktop. Downloads Obsidian. Creates a local vault. Drags her old .md folder inside. Opens Claude Code, pastes Karpathy's wiki prompt
the agent reads everything overnight, wires up her transformer notes to her mom's recipes to shower-thoughts she'd forgotten she wrote
half those notes were written in 2022 and hadn't been opened since
asks Claude "what was I thinking about attention a year and a half ago" and gets back three links to her own files
you're reading this on a laptop where the same 4 steps would work before you finish your coffee
The CEO of Obsidian just dropped 5 free Claude Code skills almost nobody knows exist.
The repo is called Obsidian Skills. Built by Steph Ango, the man who runs Obsidian itself.
5 files. That’s the whole thing.
But they fix the one problem nobody talks about: Claude Code doesn’t understand Obsidian. It sees your vault as a folder of random .md files.
These skills teach it the real system. Wikilinks. Backlinks. Canvas. Bases. Obsidian-flavored markdown.
After that, your vault stops being storage and starts being a workforce:
1.Dump 200 messy notes in. Claude links them, tags them, builds the structure while you sleep.
2.“Turn this week’s highlights into 3 post drafts.” It pulls from your notes. Your ideas, your voice, 10x the output speed.
https://t.co/G6U8DVak24 prompt → a full Canvas board mapping a project you’d spend 2 hours building by hand.
This is the exact pipeline behind the “faceless second brain” accounts: notes in, content out, daily, on autopilot.
Setup takes 2 minutes:
clone the repo from GitHub
drop the skills into your Claude Code skills folder
open Claude Code inside your vault
No plugins. No API keys. No paid tier.
Gurus charge $500+ for “AI second brain” courses built on less than these 5 free files.
The CEO gave away the manual. Most people will keep buying the course.
AI is bringing centuries of cultural heritage to life in Xi'an. The BoGuan Large Model, developed by the Shaanxi Culture Industry Investment Group, Huawei, and local partners supports digital IP creation, AI travel agents, cultural heritage preservation and drama productions.
A €10M deposit pool (8% deposit bonus for MiCA) drained on record demand.
@OKX's answer: 5x it to €50M.
That tells you two things:
- EU capital is rotating towards regulated venues, fast. OKX is the first global exchange with a full MiCA licence, and this is what that first-mover advantage looks like in real time.
- Exchanges are in a full-on war for European users right now, and that competition pays YOU.
The deal: flat 8% deposit bonus, from €10 all the way up to €250k. No tiers, no lottery. Park €10k, take home €800.
The crazy part is, the migration is happening whether you join it or not. I sure do, and while I'm doing it, might as well get paid for it 😄
The Evolution of the Agentic CPU Thesis
1/ Been going through sell side coverage on agentic CPU demand, 13 reports since April.
2/ TAM estimates are splitting into two camps. Lower, $71 to 120bn (BofA April, AMD's own pre June guidance). Upper, $170bn plus, where three independent methods are now landing, UBS via expert calls, BofA via workload segmentation ($30bn traditional plus $70bn AI head node plus $70bn agentic standalone), GS via a practitioner spend ratio framework.
3/ The mechanism is getting concrete. BofA's Rise of the Agents puts AMD Venice (256C, TSMC 2nm, 3Q26) at 3.30x the performance of NVDA Vera (88C). Turin is already shipping at 2.37x Vera. The abstract "agentic needs way more CPU" thesis now has actual products with core counts and benchmarks behind it. Just look at Vera, guiding $20bn standalone for 2026.
4/ The ARM call: It's funny seeing how fast research sentiment can change. BofA had ARM at roughly 25% unit share, rated Neutral, in April. Well... by June the same analyst had ARM at 50% value share, majority of the market by revenue. Mechanism is that ARM custom silicon (Axion, Graviton, Vera) carries 2 to 3x the ASP of merchant x86, so unit share undercounts revenue share materially. UBS independently models roughly 50% revenue share too.
5/ There is an Intel disagreement between houses. GS has a higher PT ($150) than BofA's Buy ($135) but rates it Neutral. Both houses model Intel growing. The dispute is entirely about multiple, GS says at roughly 19x 2030 base EPS the stock needs 18A to execute and consensus to migrate to the bull case. BofA says EMIB packaging is a moat that justifies the re rating now. BofA's own Rise of Agents framework actually shows AMD at slightly higher 2030 revenue share than Intel, 26% vs 24%.
6/AMD IR called their own >$120bn guidance conservative point in time in a June sell side call, and floated CPU to GPU going past 1:1 as agent fan out scales. They sell both sides of that ratio, sure. But they also said ROCm adoption friction is dropping, customers standing up AMD environments independently now, and MI450 has visibility out to CY28. The incentive is real. So is the customer data they'd be seeing that we're not.
7/ Nobody's using the same units. Headnode chip ratios, cores per GPU, attach rates, now a dollar spend ratio too. All pointing the same direction but none converts cleanly into the others, which makes it hard to stress test any single number. The one thing they do share, the directional move is toward parity or past it from every framework simultaneously.
8/ The supply constraint cuts both ways. GS's expert call says only about 50% of enterprise server demand at Dell and HPE is being fulfilled, with at least 18 months before it normalises since DRAM is going to HBM instead. That's a timing risk on when revenue shows up. It's also evidence the underlying demand is real and currently exceeding supply. BofA's memory data has DDR5 spot up 675% YoY. Nobody writing the CPU bull case has actually addressed what that does to total server economics at the margin.
9/ Two things remain structurally unaddressed across all 13 notes. Whether next gen ASICs (TPUv9, Trainium 4) fold the control plane function into the die and shrink the CPU layer everyone's forecasting off, currently ASIC racks are more CPU dense than GPU racks which is additive, but the 2028 product decisions are being made now. And how much of the global build out runs on domestic Chinese ARM silicon, which would mean the $170bn TAM is counting revenue Western vendors aren't going to see.
Many AI Agents Are Being Built To Manage Other AI Systems!
Not every AI agent is being built for customer support or trading.
An emerging category is agents supervising other agents.
• One agent performs research.
• Another validates the results.
• A third decides whether enough confidence exists to execute an action.
• A fourth monitors performance and intervenes if something goes wrong.
As agents become more specialized, coordinating them may become just as important as building them.
Why spend hundreds on videography, creators, and production when AI UGC can deliver authentic, high-converting content in minutes?
✅ No actors needed
✅ No filming setup
✅ Faster turnaround
✅ Budget-friendly
✅ Easy creative testing at scale
All you need is a product URL, and SmartUGC AI handles the heavy lifting—helping generate ad-ready UGC-style videos that would normally take days to produce.
For brands, marketers, and creators looking to move faster without sacrificing creativity, this is definitely worth trying.
Start creating AI UGC here https://t.co/NHWo3z8PBn
A brain tumor diagnosis changed everything for Melissa. Thanks to innovative treatment and a dedicated medical team, she received life-saving care. Every patient deserves the chance to benefit from the next medical breakthrough.
WATCH | KwaZulu-Natal Police Commissioner Lieutenant-General Nhlanhla Mkhwanazi says police will not stop any planned marches on Tuesday, but they will maintain public order.
Karpathy just wrote the manual for Claude + Obsidian as a real second brain.
Most vaults die the same way. A year of saved articles and highlights. None of it linked. The graph rots while it still looks impressive.
So he moved the upkeep to the model. You curate sources and ask questions. Claude files, links, and reconciles. You keep judgment. It keeps the books.
raw belongs to you and never gets edited. wiki belongs to Claude. It isn't RAG. Your sources compile once into linked pages and compound from there.
9 rules. Start with 10 sources, not 10,000.
Most people hoard notes. This turns them into a brain that maintains itself.
The AI boom is starting to reveal its real supply chain,for the public, the story is chatbots and GPUs…
But behind that are memory chips, EUV machines, advanced packaging, power contracts, cooling systems and data-center capacity…
SK Hynix going for a major U.S. listing is a reminder that AI is no longer just a software story, It is an industrial story…
The winners will be the companies building everything intelligence needs to exist.
Watch Cisco Cloud Control in action.
Discover the unified platform built for humans and AI agents to run critical IT infrastructure together, enabling customers and partners to build their own apps and agents in natural language, and extending to third-party tools.
The middle class is definitely gone. You’re either wealthy or a $1,000 repair from financial stress.
But most Americans put themselves in this position by living above their means. I know that will ruffle some feathers but it’s 100% true.
Breaking news tonight on my triggered podcast
Sen Bernie Moreno says Gavin Newsom will be subpoenaed to appear at joint hearing with Colombian election officials , to explain why Colombia can check voter ID and count votes quickly but California can’t
Love it!!!
At around 8:30 PM today, will be speaking at the #RepublicSummit2026. I will be sharing my views on India's strides across different sectors and our collective endeavour to realise the dream of a Viksit Bharat.
@republic