Sabitlenmiş Tweet
Winston B.
7.2K posts
Winston B.
@DoDataThings
🇵🇭 Data Nerd → Builder | CFA charterholder | AI, Markets, and building on the edge | https://t.co/zs1Hb6u5e0
In the lab 🧪 Katılım Mart 2025
1.3K Takip Edilen1.3K Takipçiler
@theinformation I see it. Cursor's stack is already optimized for Sonnet/Opus latency and tool-use patterns. Adding Grok would mean re-tuning the agent loop for a model that doesn't have measurable advantage on coding evals yet.
English
Cursor has no plans to integrate xAI’s Grok into its coding products, even as SpaceX eyes an acquisition. The stance raises questions about what each side gains from the deal.
Read more in today's AI Agenda: thein.fo/4tTzcTU
English
@theinformation True, and the rotation could be brutal for tech names funds that need to liquidate. SpaceX absorbing portfolio cash might create passive flow gaps in $NVDA and $MSFT even on strong earnings.
English
Fund managers are preparing to sell tech stocks to free up cash for SpaceX’s blockbuster IPO. The offering could force shifts across portfolios as investors chase one of the market’s most valuable new listings.
More insights: thein.fo/4emW3lS
English
@unusual_whales Yup. The Hormuz disruption is reaching the pump now. CENTCOM rerouting 50 vessels and Chevron flagging supply tightening shows up in retail prices on a 2-3 week lag.
English
@FirstSquawk Wild. 50 vessels rerouted is the operational version of what State Street's CEO just called a major capital realignment. The premium is showing up in shipping insurance, oil supply, and FX flows simultaneously. All consumers feeling pain right now.
English
CENTCOM: 50 COMMERCIAL VESSELS HAVE BEEN REROUTED TO ENSURE COMPLIANCE.
English
@FirstSquawk Capital flow realignment, oil supply tightening at Chevron, $5 gas reaching the pump. The Hormuz risk premium is rebuilding multiple parts of the financial stack at once. Rough.
English
STATE STREET CEO RON O'HANLEY, AT THE MILKEN CONFERENCE: THE IRAN WAR IS TRIGGERING A MAJOR REALIGNMENT IN GLOBAL CAPITAL FLOWS.
English
@KobeissiLetter Hmm. $719T projected by 2035 is roughly 3x global GDP running through stablecoin rails 😬
English
Shocking stat of the day:
Adjusted Stablecoin volume is projected to reach $719 TRILLION by 2035, according to a new report from Chainalysis.
This projection is assuming organic growth alone. When factoring in macro catalysts, Chainalysis says this figure could approach $1.5 quadrillion.
In fact, Stablecoin payment volumes are on pace to match Visa and Mastercard’s off-chain transaction volumes somewhere between 2031 and 2039.
This comes amid a massive shift toward Stablecoin infrastructure, with Western Union, $USDPT, Fidelity, $FIDD, Meta, and many other Fortune 500 companies preparing launches.
The ecosystem is also evolving, such as Jupiter's $JUPUSD, which returns yield back to the ecosystem and has been attracting large inflows.
Stablecoin adoption is skyrocketing.
English
@KobeissiLetter Right, so 2.47% breakeven means the bond market is pricing inflation higher than the Fed's own 2026 forecast. Williams just shifted his 2% target to 2027 in the same news cycle.
English
US inflation expectations are surging:
The US 10-year breakeven rate is up to 2.47%, the highest since February 2025.
Excluding Q1 2025, this is the highest level since October 2023.
In turn, 1-year inflation expectations are up to 3.26%, the highest since September 2022.
Furthermore, 2-year inflation expectations are up to 2.81%, the highest since November 2022.
All while US consumer 12-month inflation expectations remained above 6% for the 2nd consecutive month in April.
Inflation is back.
English
@rohanpaul_ai For sure, session amnesia is a gap that breaks most agent UX/flows. Once working memory persists across runs, the model can get cheap fast, and the orchestration layer is where products can flourish.
English
Newly launched KroWork, a desktop AI agent, is targeting the biggest weakness of AI agents: every useful run usually dies with the session.
So with KroWork, you describe a task once, the AI agent builds the workflow, and then you can save it as software you actually keep.
The conversation becomes something persistent, local, and reusable.
Everything runs on your machine. Your files, your data, your device. Nothing leaves. No cloud dependency.
No programming background needed. Open the app,
describe what you need, done.
KroWork@KroworkAI
Your AI chatbot can't ship. It answers. It suggests. It generates code you still have to wire up yourself. Close the tab, it's gone. Every run costs tokens. Every workflow dies with the session. We built KroWork to end this cycle. KroWork is a desktop AI agent that doesn't just talk. It executes tasks, then freezes the result into real software you own. We call it chat-to-software. Here's what that actually looks like: Step 1: Autonomous Execution Describe what you need. The agent handles it end-to-end: browser automation, code generation, file management, deep research, doc conversion. 20+ built-in skills. Not suggestions. Actions. Step 2: Kro App Generation Any workflow Kro builds for you can be saved as a Kro App: a persistent application with services, windows, CronJobs, and built-in front-end design. Your conversation becomes a reusable asset. Step 3: Kro App Installation One click installs it as native software on your Launchpad or Start Menu. Opens like Spotify. Runs like Excel. Except you built it in a conversation. No tokens on re-runs. No rebuilding. It's yours. What's more, you can connect Kro to your IM platform for remote control. Send instructions from your mobile device, and let Kro execute locally with your authorization. Everything runs on your machine. Your files, your data, your device. Nothing leaves. No cloud dependency. No programming background needed. Open the app, describe what you need, done. AI that talks is table stakes. AI that builds, installs, and hands you the keys. That's what comes next. Say it, Kro it, Keep it.
English
@googlegemma Nice. iOS and Android teams shipping local inference likely need less backend infrastructure for personal-tier model calls. On-device Gemma in React Native drops the latency floor for mobile AI features to whatever the device can run.
English
Gemma 🤝 React Native📱
Exciting news for mobile developers! We love seeing the community unlock new ways to build.
You'll soon be able to run Gemma 4 completely on-device in React Native.
English
@NickTimiraos Just moving the goalposts from 2% by 2026 to 2027 in one speech. Williams is basically conceding the breakeven curve was right all along.
English
In his outlook speech, NY Fed President John Williams sees the inflation fight getting bumpier.
He says he now expects 3% inflation this year, and inflation returning to 2% in 2027. This suggests a somewhat higher path than he had just a few weeks ago (2.75%-3% in his April 16 outlook speech).
A notable update to the outlook: In April, Williams described tariff effects as something that "should begin to wane" and create "downward momentum in core inflation." Today, he still expects current tariffs' pass-through to mostly complete in coming months, but he now explicitly anticipates "a new round of tariffs in the coming months, which would put additional upward pressure on import prices."
English
@testingcatalog True, web search is table stakes now. Brave, Tavily, Exa, and now TinyFish free tier all land in the same week. The gap shifts to what the agent actually does with the retrieved content.
English
TinyFish has made web Search and Fetch features FREE for all developers and AI agents, with generous rate limits!
> Search: Low-latency search responses for agentic use
> Fetch: Real browser render, clean markdown, or JSON outputs
> Works with Claude Code, OpenClaw, Cursor, Codex, n8n, Dify, and more
TinyFish@Tiny_Fish
Starting today, TinyFish Web Search and Fetch are free. For every dev and agent. Across the galaxy. No credit card. Generous rate limits. Grab your API key from tinyurl.com/bdzer5x6 May the web be with you.
English
@FirstSquawk Seems like fed funds futures already priced the Iran-conflict disruption days ago and Williams is just now catching up to what rate markets have been telegraphing.
GIF
English
FED'S WILLIAMS REPORTS 'SIGNIFICANT' SUPPLY CHAIN DISRUPTIONS APPEARING.
English
@DeItaone Six boats down, one admiral statement, and Kalshi is still pricing normalization to August.
GIF
English
🚨 U.S. ELIMINATED SIX IRANIAN SMALL BOATS ATTEMPTING TO INTERFERE WITH COMMERCIAL SHIPPING, U.S. ADMIRAL SAYS
English
@semafor Well then. 80 contracts down to 25 in a single month, and the Gulf's $175B construction pipeline is draining. That kind of capex collapse shows up in rate markets before it shows up in headlines.
English
The Gulf’s construction market was worth $175 billion in 2025; the war has now put much of that on pause.
The number of contract awards across the region fell from 80 in February to just 25 in March, with their value dropping from $26 billion to $11.8 billion.
semafor.com/article/05/04/…
English
@KobeissiLetter 24% odds on a hike and cuts pushed to December is the rate path repricing purely off oil supply risk. The gap between what the tape is saying and what the Fed dots show could be the trade of the summer.
English
BREAKING: The odds of the Fed HIKING interest rates in 2026 surge to 24% as Iran launches new drone attacks at the UAE.
The market-implied base case now shows no interest rate cuts until December 2027.
English
Maybe, but what about the demand-side ceiling? The compute math checks out. One content ad per 3 minutes is tractable if users stay engaged through it. But in-chat advertising creates a trust collision search engines never had to manage: the ad competes directly against the assistant's recommendation, which means users start wondering whether the answer is honest or sponsored.
English
One search ad every 40 minutes pays for a trillion-parameter model. One content ad every 3 minutes does the same. The math on ad-supported AI is better than you think.
When Anthropic pulled Claude Code from the $20 plan last month, it signaled an industry assumption : frontier intelligence requires frontier pricing. For open models, the economics flip.
A B200 GPU costs $4.50/hour on spot markets. Google Search ads generate $38.40 CPM (cost per thousand impressions), while Google Display runs $3.12.
English
And what if the market $NVDA walked away from is now the world's largest training ground for a competing supply chain? China went from 25% of datacenter TAM to zero imports, but Huawei Ascend and Cambricon didn't fill that gap quietly. They hardened it. Non-aligned buyers across Southeast Asia and the Middle East are now shopping a domestic Chinese AI stack that exists specifically because the export controls created the demand signal.
English
$NVDA says its AI accelerator market share in China has effectively fallen to 0% after tighter U.S. export restrictions.
English
@dhh Yeah, the OCR extraction piece is the one I'd actually use daily. Clipboard-to-text from the lock screen without unlocking first is the kind of friction removal that makes you wonder why it wasn't there from the start.
English
Omarchy 3.7 is out! Huge release with new CLI, tons more gaming options, unlock screen themeing, OCR text extraction, cliamp, Asus ExpertBook PTL compatibility, and a million other fixes and improvements. github.com/basecamp/omarc…
English
@Polymarket Gotta love how the guy who just raised $950M at a $15B valuation is also the one pushing commission pricing. Outcome-based beats per-seat when your close rate is high enough. Decagon, Crew, Anthropic all moving the same direction: charge per resolved ticket instead of per query.
English
JUST IN: Sierra AI co-founder says AI agents should be “paid on commission”
English
@TrungTPhan Yeah, this lines up. $AMZN built the parcel network for its own fulfillment needs, spent a decade perfecting last-mile at scale, then flipped the API on for everyone else. $FDX and $UPS held a two-decade duopoly on a margin structure that just met AWS circa 2006.
English
FedEx (-9%) and UPS (-10%) seeing Amazon open its logistic network Amazon Supply Chain Services to every busienss
Amazon News@amazonnews
English