@LuvsPupTech
Account based inUnited States!
About this account
- Account based in
- United States
- Connected via
- Web
X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
Ex-HealthTech entrepreneur, loves pets 🐶 and tech 🖥️, often found golfing ⛳.
London
Joined December 2023
- Tweets2.6K
- Following175
- Followers65
- Likes9.4K
TechPupJames retweeted
🚨 CRYPTO: Visa, Reap to launch stablecoin cards in 100 markets
Reap, a Hong Kong-headquartered fintech, said it is working with Visa to launch stablecoin-linked Visa credit card programs in more than 100 markets.
The partnership will extend Reap’s card-issuing infrastructure beyond Asia and Latin America to Europe, the Middle East, and Africa.
The companies said fintechs, businesses, and platforms would be able to offer Visa card programs that use stablecoins for funding, repayment, and cross-border spending, while using Visa’s network.
They also said they would explore new settlement and payment options, including agentic commerce.
Reap added that it is part of Visa’s stablecoin settlement program in Asia Pacific.
The move is part of Visa’s strategy to enable stablecoin-linked cards and follows another effort in which it partnered on stablecoin payments.
In Europe, expansion could be shaped by the European Union’s Markets in Crypto-Assets (MiCA) rules, which require authorization, governance, reserve backing, transparency, and redemption standards for many fiat-referenced stablecoins used in payments.
Source: Tech in Asia
#Fintech #payments #crypto @Visa
TechPupJames retweeted
SEC is working on ‘boring’ rules for Crypto. The industry needs clarity for stability.
TechPupJames retweeted
this is f*cking gold
There's an official Anthropic plugin called claude-code-setup, and almost nobody has installed it.
It reads your project and tells you which automations to switch on: hooks, skills, MCP servers, subagents, and how to set each one up step by step. One line to install: /plugin install claude-code-setup@claude-plugins-official
It's the onboarding the one-person companies wrote by hand. Save it ↓
TechPupJames retweeted
Anthropic's first advice for Opus 5.5 fits in three lines. Only the first one changes how you work.
"Hand over a whole task. Define 'done' and when to check in." That's a job description in two sentences: the outcome, the finish line, the review point. Write it once as a file and it's a skill.
Most people will read the second line, delete "think carefully" from their prompts and call it an upgrade. The first line is where the hours come back.
The job descriptions one-person businesses run on ↓
TechPupJames retweeted
🚨BREAKING: the open-source project quietly saving enterprises millions in GPU spend.
LMCache: 11.8K stars, 124K downloads a month, built with contributors from NVIDIA and Redis.
The problem:
your LLM recomputes the entire context from zero on every message system prompt, history, retrieved docs. Thousands of GPU cycles, wasted, per request.
LMCache:
fixes it by keeping the KV cache alive and shareable across instances, offloaded down a tiered stack GPU, then CPU DRAM, then disk, then network storage.
That unlocks:
1. Instant RAG cache a PDF once, queries start instantly
2. Disaggregated prefill heavy work on H100s, decode on L4s
3. Shared context compute once, serve every user
Result:
3-10x delay savings on multi-round QA.
Plugs into vLLM and SGLang. Apache-2.0. pip install lmcache.
Stop paying your GPUs to redo homework they already finished.
repo in the comments.
TechPupJames retweeted
Thank you for the inclusion, @Fabriziobustama!
Love this — everyday infrastructure becoming a connected smart-city platform.
#SmartCities #IoT #5G #FutureMobility
A bus stop is about to become much more than a place to wait. 🚌⚡
This futuristic smart shelter concept combines:
📡 5G connectivity
🖥️ Real-time digital information
🔋 Wireless charging
📢 Interactive advertising
🌃 Smart-city infrastructure
☀️ Potential solar integration
The interesting part isn’t the shelter itself.
It’s the idea that everyday urban infrastructure could become a connected digital platform — delivering mobility data, communication, charging and public services in one place.
The future of transportation may be built one bus stop at a time.
Would you want this in your city? 👀
#SmartCities #FutureMobility #Innovation #Technology #Tech #AI #IoT #5G #Bigdata #Robotics
Cc: @jblefevre60 @Nicochan33 @pierrepinna @RosyCoaching @kalydeoo @Ym78200 @enilev @FrRonconii @CurieuxExplorer @Khulood_Almani @mvollmer1 @JoannMoretti @margaretsiegien @Shi4Tech @anand_narang @Eli_Krumova @HaroldSinnott @bimedotcom @EvanKirstel @Hana_ElSayyed @sallyeaves @SpirosMargaris @ipfconline1 @LaurentAlaus @Analytics_699 @Jeffkagan @AndresVilarino @efipm @FMFrancoise @MHcommunicate @Corix_JC @Archonsec @JolaBurnett @JeromeMONANGE @mary_gambara @drsharwood @smoothsale @sulefati7 @debashis_dutta @dinisguarda @devbhatt72 @terence_mills @BetaMoroney @mdrechsler @MarshaCollier @avrohomg @GeekOnTheLoose @NevilleGaunt @PamMktgNut @arlenenewbigg
TechPupJames retweeted
Google just dropped a 1-hour course on full AI engineering: from 1 agent to Loops and Graphs
7:02 - build your first agent
13:34 - what agents can do
23:08 - context engineering
32:56 - graph engineering
54:11 - full autonomous system
one agent saves you an hour. a thousand replaces the team.
Prompts → Agents → Loops → Graphs
most people will stop after the second timestamp and call it a system
he spends the back half on the part that is still true next year
same model, same tokens, completely different week
watch it today
the full guide on loops and graphs is below, save it while it is still early ↓
TechPupJames retweeted
So GPT-6 Sol only scores 48 on the AA intelligence index, 1 point more than GPT-5.6 Sol..
Luna sits at 37, exactly the same as 5.6 Luna.
The intelligence didn't move, the only real change is price, both got cut ~in half which is a great thing.
"GPT-6" isn't a new generation of brains, it's a price cut with a new number on it.
Meanwhile Opus 5.5 moved the ceiling to 58, beat Astra, and still dropped 20% vs Opus 5.
Please welcome GPT-6 Sol and GPT-6 Luna to the GPT-6 universe.
GPT-6 Sol and Luna build on the advances behind GPT-6 Astra, bringing much of its strengths into faster and more affordable models to support work at scale.
We’ve also made caching and inference more efficient, and we’re passing the savings directly to you: 50% lower API prices for Sol and Luna compared with GPT‑5.6 promotional pricing.
TechPupJames retweeted
Ripple Did Not Issue a Token Looking For a PURPOSE - It Built a GLOBAL Financial Ecosystem Around The Worlds Greatest Need: TRUSTED, INSTANT LIQUIDITY.
With relationships spanning central banks, global banks, fintechs, exchanges and payment providers, Ripple already stands at the intersection of legacy finance and emerging DLT infrastructure. Ripple supplies the institutional enterprise layer. XRPL supplies the open, decentralized settlement ledger. RLUSD supplies regulated digital-dollar stability. XRP supplies scarce, neutral, jurisdictionless bridge liquidity - without requiring any nation, bank or blockchain to surrender sovereignty.
As money, assets and ownership move toward 24/7 tokenization and atomic settlement, few ecosystems can match this combination of global reach, regulatory positioning, interoperability, speed, liquidity and proven infrastructure. The future will not belong merely to the blockchain with the loudest following - but to the ecosystem capable of connecting everything. @Ripple, @XRPLF, XRP, RLUSD and @USTreasury may already be standing exactly where the entire financial world is heading.
TechPupJames retweeted
🇺🇸 JUST IN: CFTC Chairman Selig says US markets must prepare for "mass tokenization" as blockchains and AI are "adopted at scale."
Selig told the US Treasury Market Conference that "the next decade will likely bring more change to financial markets than the previous several decades combined."
The CFTC expanded eligible collateral to include stablecoins in February and is looking to drive further adoption across exchanges and clearinghouses.
TechPupJames retweeted
A scammer emailed @Revolut from a real government domain. Staff sent over 680 customers' passports, addresses and selfies.
@FT and @WSJ have it. The files are on @telegram with a ransom demand.
Presented by @formancehq
TechPupJames retweeted
opus 5 writes code, gpt-6 astra handles chat, gemini 3.1 does vision, perplexity runs web search, and llama 4 answers private queries locally, all in one desktop window while sam altman keeps charging $20 a month for chatgpt plus, dario amodei charges $20 for claude pro, and adam d'angelo charges $20 for poe premium separately
cherry studio ships as a native installer for windows, mac, and linux with no docker or cli setup needed:
download the installer, paste your api keys, and 300+ pre-configured assistants switch between openai, anthropic, gemini, deepseek, perplexity, poe, ollama, and lm studio in one conversation
it's called cherry-studio and here's what cherryhq quietly shipped past 52k stars:
52k stars, agpl-3.0 desktop client for windows mac and linux with native support for openai, anthropic, gemini, perplexity, poe, ollama, and lm studio in one chat, plus 300+ pre-configured assistants, multi-model simultaneous conversations, mcp tool calling, and webdav sync, matching the features chatgpt plus, claude pro, and poe premium bundle for $60 a month combined
what actually happens when you open cherry studio on your desktop:
lets you swap between openai, anthropic, gemini, deepseek, perplexity, poe, ollama, and lm studio without opening a new tab or copying an api key twice
-> boots 300+ pre-configured assistants for coding, writing, translation, and document analysis with drag-and-drop sorting
-> spawns mcp servers so any model gets tool calling for files, terminals, browsers, and custom apis
-> renders full markdown, mermaid diagrams, code syntax highlighting, and inline pdf, office, and image previews
-> syncs conversations and configs to webdav, exports to pdf, and runs fully offline with local ollama models
sam altman charges $20 a month for chatgpt plus. dario amodei charges $20 for claude pro. adam d'angelo charges $20 for poe premium. one team in china shipped the desktop client that replaces all three for the price of an openrouter api key
TechPupJames retweeted
50 Claude Skills every CFO should know about
Most finance teams build zero Skills.
They rewrite the same prompt from scratch
every single month
I help finance teams build their first skill in my AI CFO Masterclass tomorrow
👉 Comment "Skills" for a free invite!
THIS IS F**KING GOLD
I found Laya, which now runs natively on Apple Silicon
Not via a cloud API
Not via PyTorch
Not by generating JSON token by token
laya-mlx runs open-weight decision models locally using MLX
It won't write code or a post for you
But in a single pass, it can solve the kinds of tasks for which we usually wastefully query a large LLM:
▸ where to route a ticket
▸ how urgent it is
▸ whether the message contains a jailbreak
▸ which tool to call
▸ whether to skip the next action
On an M3 Max, the author shows:
> 421M English Laya: P50 13.42 ms
> 322M multilingual: P50 7.39 ms
> Peak MLX memory: 943.6 MiB and 687.6 MiB
And this isn't just some abstract "JSON benchmark.
The repository includes a Snake demo where every move gets a new decision from the model locally
In an optimized run: 75.4 moves per second over 2,400 moves
But there are important limitations:
▸ Apple Silicon only, macOS 14+
▸ It’s an independent MLX port, not an official release from the Laya team
▸ Model context is 512–1024 tokens
▸ It’s a decision layer, not a replacement for Codex, Claude, or a standard LLM
The most interesting part isn't the Snake game
It’s that you can keep a large LLM in your app for complex work while offloading quick `choose / score / yes-no` decisions to a local layer with millisecond latency
Fewer API calls
Less data sent externally
Faster execution for chains where one decision blocks the next
TechPupJames retweeted
Anthropic engineer:
"99% of people use Claude Code like Google, and only 1% are running swarms of self-learning Claude agents
I'm running 100+ agents in a loop. I have Chief agent, PM agents - they manage the whole team"
in a 30-minute workshop, an Anthropic engineer revealed how to get max value from Claude Code at min. cost
this is worth more than another $500 vibe-coding course
watch today, then read how to build self-improving agentic systems with Fable in the article below
TechPupJames retweeted
Mixture-of-Experts (MoE) models have primarily concentrated on introducing sparsity within the Feed Forward Network (FFN) layers. MoVA extends this concept by applying sparsity to the value computation mechanism within the attention layers.
This approach results in 4 billion active parameters per token, achieves a score of 25 on the Artificial Analysis Intelligence Index, and does not incur additional KV-cache costs when compared to standard Grouped-Query Attention (GQA).
It will be interesting to observe the future development and adoption of this technique.
K2-Horizon-36B-A4B scores 25 on the Artificial Analysis Intelligence Index, matching models with over 20× the total parameters while using 4B active parameters per token.
These capabilities come from our new architecture MoVA (Mixture-of-Value Attention), which incorporates MoE-based sparsity into the compute of value vectors in multi-head attention.
It opens a second axis for scaling sparsity in an LLM, beyond MoE in the FFN module. Importantly, MoVA enjoys the following advantages:
• Simple and compatible with efficient attention algorithms, such as flash attention, GQA, and sparse attention
• No additional KV cache cost comparing to standard GQA
K2-Horizon-36B-A4B available at: huggingface.co/IFM/K2-Horizo…
TechPupJames retweeted
Most people use Claude for writing.
That’s only a small part of what it can do.
The real opportunity is turning Claude into a specialized skill set for your entire workflow.
This list of 50 Claude skills shows just how far you can take it:
From development & tech:
→ Debugging
→ Testing
→ Git workflows
→ Security checks
→ Code quality
To research & knowledge:
→ YouTube research
→ Article extraction
→ Content research
→ PDF analysis
→ Brainstorming
Then productivity & automation:
→ File organization
→ Invoice management
→ Asset generation
→ Automated workflows
And it goes even further.
You can build systems for:
→ SQL & Excel
→ Dashboards
→ A/B testing
→ Landing page copy
→ Ad creatives
→ Sales funnels
→ Email campaigns
→ Brand voice
And for creators:
→ YouTube scripts
→ Podcast outlines
→ Video editing
→ Storytelling
→ Content repurposing
The interesting part isn’t having 50 skills.
It’s realizing that one AI can become a collection of specialized workflows instead of one general-purpose chatbot.
Think of it like this:
Claude + the right skills = AI team for different jobs.
You don’t need to learn all 50.
Pick the 2–3 skills that solve your biggest bottlenecks, build them into your workflow, and expand from there.
Save this list. Your next AI workflow might already be on it.
Follow @AamirAnsar94694 for more practical AI, tech & automation insights.
#AI #Claude #AITools #Automation #AIWorkflows #GenerativeAI #Productivity #FutureOfWork
TechPupJames retweeted
Damn. I’m late to the party. Plane just arrived in SF: Grok 4.7 looks super amazing!
Not only is it competitive to sota and even surpasses fable 5.1 in some benchmarks, it’s also so much cheaper than GPT and Fable 5.1!
Well done spaceX!! Will write about it a bit more later on
TechPupJames retweeted
I was too quick and excited when landed in SF. As most of you pointed out the real deal is actually MiMo-2.6!
Better than Grok 4.7, much cheaper and holy moly is china back. That is the real surprise!
I need to check that out. Message to chubby: next time checking x properly before posting :D
