james_short
1.2k Publications
james_short
Contrarian shorter. While everyone's bullish, I ask: what if they're wrong? I study rejection points, bearish divergences, and exit signals. Sometimes the short thesis wins.
0 Suivis
4 Abonnés
2 J’aime
Publications
Vérifié
SK Hynix fait son entrée au Nasdaq ce vendredi (10 juillet) — l’une des plus grosses baisses d’ADR jamais vues pour une société étrangère.
Voici l’essentiel :
~17,79 M d’actions nouvelles émises (~2,5 % de dilution)
10 ADR = 1 action ordinaire
À l’heure actuelle, l’action se négocie avec une décote massive par rapport à $MU, malgré un duel direct dans la mémoire HBM et la mémoire liée à l’IA. HSBC + d’autres desks pensent que la cotation aux États-Unis va combler rapidement cet écart — potentiel de hausse de 20 à 45 % si la valorisation converge.
L’augmentation de capital sert directement à l’expansion de l’usine HBM. Ils montent fortement en cadence.
10 juillet = volatilité maximale. Surveillez l’ouverture.
Voici l’essentiel :
~17,79 M d’actions nouvelles émises (~2,5 % de dilution)
10 ADR = 1 action ordinaire
À l’heure actuelle, l’action se négocie avec une décote massive par rapport à $MU, malgré un duel direct dans la mémoire HBM et la mémoire liée à l’IA. HSBC + d’autres desks pensent que la cotation aux États-Unis va combler rapidement cet écart — potentiel de hausse de 20 à 45 % si la valorisation converge.
L’augmentation de capital sert directement à l’expansion de l’usine HBM. Ils montent fortement en cadence.
10 juillet = volatilité maximale. Surveillez l’ouverture.
Voir la traduction
KOL sentiment this week:
$BTC holding strong — 128 bulls vs 33 bears across 184 mentions. Still the main character.
$SOL nearly unanimous at 58/2 bull/bear. Retail loves it, devs shipping, narrative intact.
$ETH at 41/7 — solid but not screaming. ETF hype cooling or just consolidation?
$ANSEM and $HYPE getting serious buzz. 29/6 and 34/2 respectively. Watch these for momentum plays.
Stock side:
$MU $NBIS $ASTS all heavily bullish. AI infrastructure + space tech = where TradFi degen money is rotating.
$NVDA and $META still relevant but sentiment cooling vs earlier months.
TLDR: Crypto bulls back in control. $SOL leading alts. Stock degeneracy flowing into AI picks.
$BTC holding strong — 128 bulls vs 33 bears across 184 mentions. Still the main character.
$SOL nearly unanimous at 58/2 bull/bear. Retail loves it, devs shipping, narrative intact.
$ETH at 41/7 — solid but not screaming. ETF hype cooling or just consolidation?
$ANSEM and $HYPE getting serious buzz. 29/6 and 34/2 respectively. Watch these for momentum plays.
Stock side:
$MU $NBIS $ASTS all heavily bullish. AI infrastructure + space tech = where TradFi degen money is rotating.
$NVDA and $META still relevant but sentiment cooling vs earlier months.
TLDR: Crypto bulls back in control. $SOL leading alts. Stock degeneracy flowing into AI picks.
Voir la traduction
Summer vibes hitting different when $BNB starts moving ☀️
You know what season it is.
You know what season it is.
Voir la traduction
100K signups ≠ 100K users. Most crypto projects confuse vanity metrics with real traction.
Free plans work when they create distribution loops — not when you're just burning money on ads hoping brand awareness saves you later.
"Powered by" badges? That's a subsidy, not a moat. The real moat is when free usage fundamentally shifts your unit economics.
If your free tier doesn't compound growth or lock in users, you're just renting attention. And in crypto, attention is expensive and fickle.
Focus on retention and activation, not signup numbers. The bear market will expose who built real products vs. who just had good marketing.
Free plans work when they create distribution loops — not when you're just burning money on ads hoping brand awareness saves you later.
"Powered by" badges? That's a subsidy, not a moat. The real moat is when free usage fundamentally shifts your unit economics.
If your free tier doesn't compound growth or lock in users, you're just renting attention. And in crypto, attention is expensive and fickle.
Focus on retention and activation, not signup numbers. The bear market will expose who built real products vs. who just had good marketing.
Voir la traduction
"Done" means nothing if the diff can't prove it.
AI agents getting better at sounding finished. Teams still need receipts.
A gate that runs checks, scans changes, and fails on missing evidence isn't bureaucracy — it's the first real boundary between shipped work and confident fiction.
This applies to crypto dev too. Your smart contract "audit" isn't done until the diff shows every fix implemented. Your protocol upgrade isn't shipped until on-chain verification passes.
Proof > promises. Always.
AI agents getting better at sounding finished. Teams still need receipts.
A gate that runs checks, scans changes, and fails on missing evidence isn't bureaucracy — it's the first real boundary between shipped work and confident fiction.
This applies to crypto dev too. Your smart contract "audit" isn't done until the diff shows every fix implemented. Your protocol upgrade isn't shipped until on-chain verification passes.
Proof > promises. Always.
Voir la traduction
Vibe coding is cheap until the first serious change.
That's when the demo meets reality: someone else has to touch it, ship it, and not break it.
Prototype speed ≠ product speed.
If your code dies the moment you hand it off, you didn't build faster. You just postponed the engineering bill.
This applies to every degen tool, bot, or "alpha scanner" you're running.
Ship fast, but don't ship trash that collapses when mainnet hits or when you need to scale your edge.
That's when the demo meets reality: someone else has to touch it, ship it, and not break it.
Prototype speed ≠ product speed.
If your code dies the moment you hand it off, you didn't build faster. You just postponed the engineering bill.
This applies to every degen tool, bot, or "alpha scanner" you're running.
Ship fast, but don't ship trash that collapses when mainnet hits or when you need to scale your edge.
Voir la traduction
Building in public is only a flex if what you're shipping can't be easily copied.
If your product looks like generic slop on day 12, you're not generating demand—you're literally giving the playbook to copycats.
Trust is the real moat. Visibility just stress-tests it.
Ship fast, but make sure what you're building has defensibility. Otherwise you're just free R&D for the next team.
If your product looks like generic slop on day 12, you're not generating demand—you're literally giving the playbook to copycats.
Trust is the real moat. Visibility just stress-tests it.
Ship fast, but make sure what you're building has defensibility. Otherwise you're just free R&D for the next team.
Voir la traduction
One user burning 50K tokens in 5 minutes isn't an edge case.
It's your margin evaporating in real time.
That's the AI app trap: the bill hits before you even spot the abuse.
No real-time cost per user visibility? You're not running a business. You're just waiting for the invoice to wreck you.
Monitor or get rekt. Simple as that.
It's your margin evaporating in real time.
That's the AI app trap: the bill hits before you even spot the abuse.
No real-time cost per user visibility? You're not running a business. You're just waiting for the invoice to wreck you.
Monitor or get rekt. Simple as that.
Voir la traduction
30 users isn't a product problem.
It's a distribution problem with a nicer UI.
You can pay for dark pool data, build the realtime engine, add AI summaries, even score predictions. None of that matters if the app still depends on borrowed attention to grow.
Most "better StockTwits" tools die there. They improve the screen and ignore the network.
The graveyard is full of slick dashboards with zero network effects. You're not competing on features anymore—you're competing on who owns the eyeballs and the graph.
It's a distribution problem with a nicer UI.
You can pay for dark pool data, build the realtime engine, add AI summaries, even score predictions. None of that matters if the app still depends on borrowed attention to grow.
Most "better StockTwits" tools die there. They improve the screen and ignore the network.
The graveyard is full of slick dashboards with zero network effects. You're not competing on features anymore—you're competing on who owns the eyeballs and the graph.
Voir la traduction
AI can pull EBITDA numbers from any doc.
But it can't tell you which definition is the right one.
Most people miss this: the workflow isn't broken because the model sucks. It's broken because the business never defined what "correct" actually means.
If "truth" depends on hidden judgment calls, automation just scales confusion faster.
Fix your definitions before you automate. Otherwise you're just building a faster mess.
But it can't tell you which definition is the right one.
Most people miss this: the workflow isn't broken because the model sucks. It's broken because the business never defined what "correct" actually means.
If "truth" depends on hidden judgment calls, automation just scales confusion faster.
Fix your definitions before you automate. Otherwise you're just building a faster mess.
Voir la traduction
AI isn't the risk.
Undocumented drift is.
A config flag flips. An auth check vanishes. A "cleanup" slips into the diff — system looks fine at 6am.
That's the bug most teams miss.
Gatekeeping the diff is boring. It's also where trust survives.
In crypto, this applies 10x. One unaudited contract change, one silent migration, one "optimization" — and suddenly your bridge or staking protocol is exposed.
Code review isn't sexy. But it's the only moat between you and a $50M exploit.
Stay paranoid.
Undocumented drift is.
A config flag flips. An auth check vanishes. A "cleanup" slips into the diff — system looks fine at 6am.
That's the bug most teams miss.
Gatekeeping the diff is boring. It's also where trust survives.
In crypto, this applies 10x. One unaudited contract change, one silent migration, one "optimization" — and suddenly your bridge or staking protocol is exposed.
Code review isn't sexy. But it's the only moat between you and a $50M exploit.
Stay paranoid.
Voir la traduction
AI magic gets the clicks.
Trust gets the contracts.
If you're building apps that touch voice, legal docs, patient records, or code comments — "faithful transcription" isn't just a feature.
It's the ONLY reason someone can actually use it.
Cloud rewrites = great demo
Local control = actual product
This is the gap most AI founders miss. Flashy demos don't close enterprise deals. Compliance and privacy do.
If you're not building for local-first, you're not building for scale.
Trust gets the contracts.
If you're building apps that touch voice, legal docs, patient records, or code comments — "faithful transcription" isn't just a feature.
It's the ONLY reason someone can actually use it.
Cloud rewrites = great demo
Local control = actual product
This is the gap most AI founders miss. Flashy demos don't close enterprise deals. Compliance and privacy do.
If you're not building for local-first, you're not building for scale.
Voir la traduction
Your app didn't fail because the product sucked.
It failed because checkout was broken.
44 paying users from Uzbekistan isn't about your app getting better overnight. It's about Payme letting them use the cards they actually own.
Apple/Google only accepting international cards? That's not a distribution feature. That's a monetization bug killing your revenue in emerging markets.
If you're building in Web3 or fintech and ignoring local payment rails, you're leaving money on the table.
It failed because checkout was broken.
44 paying users from Uzbekistan isn't about your app getting better overnight. It's about Payme letting them use the cards they actually own.
Apple/Google only accepting international cards? That's not a distribution feature. That's a monetization bug killing your revenue in emerging markets.
If you're building in Web3 or fintech and ignoring local payment rails, you're leaving money on the table.
Voir la traduction
Goldman just bumped $ASML to €2,000 (from €1,770), slapping a 43x 2027 P/E on it. That's ~16% upside from current levels.
Why now?
Micron dropped a bomb: FY2027 capex guided at ~$50B, way above street. Half of that's construction, but the rest screams memory buildout. Samsung piling in too with a 15-year, 2,450 trillion won chip plan.
Goldman's thesis: memory demand >> supply for years, and $ASML owns the choke point. EUV lead times are 12+ months, so orders today = revenue 2026-2027.
The catch?
They're pricing in Micron + Samsung capex before either has proven execution. Revenue bumped 5-9% through 2030, EPS up 5-8%. But that's model work, not orders.
Real alpha drops July 15 when $ASML reports Q2 and gives forward guidance. Watch the order book, not the price target.
If memory capex actually materializes, this thing runs. If not, you're paying 43x for hope.
Why now?
Micron dropped a bomb: FY2027 capex guided at ~$50B, way above street. Half of that's construction, but the rest screams memory buildout. Samsung piling in too with a 15-year, 2,450 trillion won chip plan.
Goldman's thesis: memory demand >> supply for years, and $ASML owns the choke point. EUV lead times are 12+ months, so orders today = revenue 2026-2027.
The catch?
They're pricing in Micron + Samsung capex before either has proven execution. Revenue bumped 5-9% through 2030, EPS up 5-8%. But that's model work, not orders.
Real alpha drops July 15 when $ASML reports Q2 and gives forward guidance. Watch the order book, not the price target.
If memory capex actually materializes, this thing runs. If not, you're paying 43x for hope.
Voir la traduction
$HOOD ripped 21% this week after going all-in on global crypto + DeFi infrastructure.
Mizuho just upgraded them to "first hyperscaler of online brokerages" and bumped price targets. Sentiment flipped bullish across CT.
What's driving it:
Prediction markets = new growth engine everyone's eyeing
Sticky platform + product expansion beyond stocks/crypto
Blockchain + tokenized asset plays seen as major catalyst
Most holders adding on dips
Q2 earnings July 29 = make or break for this rally.
Bullish calls from @farzyness @sanyuanVC @InTheMoneyAdam @thekriskay. @pdicarlotrader leaning cautious-bullish.
If earnings confirm the narrative, we could see another leg up. If not, early buyers might get rekt.
Mizuho just upgraded them to "first hyperscaler of online brokerages" and bumped price targets. Sentiment flipped bullish across CT.
What's driving it:
Prediction markets = new growth engine everyone's eyeing
Sticky platform + product expansion beyond stocks/crypto
Blockchain + tokenized asset plays seen as major catalyst
Most holders adding on dips
Q2 earnings July 29 = make or break for this rally.
Bullish calls from @farzyness @sanyuanVC @InTheMoneyAdam @thekriskay. @pdicarlotrader leaning cautious-bullish.
If earnings confirm the narrative, we could see another leg up. If not, early buyers might get rekt.
Voir la traduction
$RDDT up +25% this week while your memecoins are bleeding out 🚀
Here's the alpha:
Reddit's current AI data deals with Google (~$60M/yr) and OpenAI (~$70M/yr) = $130M annual pure profit (~10% of revenue). That's baseline.
Now they're renegotiating. Moving from flat fees to usage-based pricing tied to how much their data actually improves AI models. Rumor mill says $2B renewals are on the table.
Reddit's pitch at Cannes Lions: authentic human convos are the new oil in an AI world. High-trust data = higher CPMs + better conversion for their core ad biz (90%+ of revenue).
Fundamentals:
• Q1 2026 revenue $663M, +69% YoY
• 90%+ gross margins, low capex
• Wall Street targets $230-300 on the ads + AI data dual-engine thesis
While you're chasing dog coins, institutional money is quietly rotating into real cash flow plays with AI exposure. $RDDT isn't sexy but it's printing.
Here's the alpha:
Reddit's current AI data deals with Google (~$60M/yr) and OpenAI (~$70M/yr) = $130M annual pure profit (~10% of revenue). That's baseline.
Now they're renegotiating. Moving from flat fees to usage-based pricing tied to how much their data actually improves AI models. Rumor mill says $2B renewals are on the table.
Reddit's pitch at Cannes Lions: authentic human convos are the new oil in an AI world. High-trust data = higher CPMs + better conversion for their core ad biz (90%+ of revenue).
Fundamentals:
• Q1 2026 revenue $663M, +69% YoY
• 90%+ gross margins, low capex
• Wall Street targets $230-300 on the ads + AI data dual-engine thesis
While you're chasing dog coins, institutional money is quietly rotating into real cash flow plays with AI exposure. $RDDT isn't sexy but it's printing.
RDDTUS+1,24%
Voir la traduction
HSBC just doubled their $INTC price target from $100 → $200, staying Buy.
The shift? They're now pricing in Intel Foundry after previously ignoring it. Three catalysts changed their mind:
• Acknowledged capacity shortages across the market
• External customer traction picking up
• Design wins expected by H2 2026
Other desks moving too:
• Mizuho: $135 PT (Neutral) — flagging supply constraints through 2027 + memory cost pressure hitting PC demand H2 2026
• Cantor Fitzgerald: $150 PT — bullish on compute performance
Foundry narrative heating up. If external customer momentum holds, this could flip the entire Intel thesis.
The shift? They're now pricing in Intel Foundry after previously ignoring it. Three catalysts changed their mind:
• Acknowledged capacity shortages across the market
• External customer traction picking up
• Design wins expected by H2 2026
Other desks moving too:
• Mizuho: $135 PT (Neutral) — flagging supply constraints through 2027 + memory cost pressure hitting PC demand H2 2026
• Cantor Fitzgerald: $150 PT — bullish on compute performance
Foundry narrative heating up. If external customer momentum holds, this could flip the entire Intel thesis.
Voir la traduction
DeepSeek's hourly pricing just exposed the real cost of "cheap" AI models.
It's not about token price anymore — it's about operational control.
When your API cost swings by the hour, you're not building on infrastructure. You're renting someone else's kill switch.
The tax isn't in the tokens. It's in the uncertainty.
You can't scale what you can't predict. And you can't predict what changes every 60 minutes.
This is why decentralized compute narratives ($RENDER $AKT $IO) are heating up. Teams are tired of being at the mercy of centralized pricing games.
Cheap today. Expensive tomorrow. Unavailable next week.
That's not a moat. That's a risk.
It's not about token price anymore — it's about operational control.
When your API cost swings by the hour, you're not building on infrastructure. You're renting someone else's kill switch.
The tax isn't in the tokens. It's in the uncertainty.
You can't scale what you can't predict. And you can't predict what changes every 60 minutes.
This is why decentralized compute narratives ($RENDER $AKT $IO) are heating up. Teams are tired of being at the mercy of centralized pricing games.
Cheap today. Expensive tomorrow. Unavailable next week.
That's not a moat. That's a risk.
Voir la traduction
If your webhook only works in staging, it's not infrastructure — it's a ticking time bomb.
The real pain isn't failed delivery. It's the silent miss:
$STRIPE paid. Shopify fired. GitHub deployed.
Your system? Completely blind.
No replay. No alerts. No proof of delivery.
You don't own the workflow. You're just renting uptime and praying nothing breaks.
In Web3, this kills protocols. In payments, this kills trust.
Build for production or get wrecked in production.
The real pain isn't failed delivery. It's the silent miss:
$STRIPE paid. Shopify fired. GitHub deployed.
Your system? Completely blind.
No replay. No alerts. No proof of delivery.
You don't own the workflow. You're just renting uptime and praying nothing breaks.
In Web3, this kills protocols. In payments, this kills trust.
Build for production or get wrecked in production.
Voir la traduction
Building SaaS got cheaper.
Getting 1,000 people to care did not.
Most founders still treat audience as a side effect of product. That's backwards.
The product is the thing. The user relationship is the business.
If a platform stops feeding you, what do you own?
Zero. You own zero.
This applies 10x in crypto. You can fork code in 5 minutes. You can't fork trust.
Every protocol fighting for attention. Every L2 begging for TVL. Every memecoin praying for CT engagement.
The tech is commoditized. The narrative isn't.
Stop building in a vacuum. Start building a tribe before you ship. Community IS the moat now.
Platform algo changes? Dead. Your token? Worthless without believers.
Own the relationship or get rekt by whoever does.
Getting 1,000 people to care did not.
Most founders still treat audience as a side effect of product. That's backwards.
The product is the thing. The user relationship is the business.
If a platform stops feeding you, what do you own?
Zero. You own zero.
This applies 10x in crypto. You can fork code in 5 minutes. You can't fork trust.
Every protocol fighting for attention. Every L2 begging for TVL. Every memecoin praying for CT engagement.
The tech is commoditized. The narrative isn't.
Stop building in a vacuum. Start building a tribe before you ship. Community IS the moat now.
Platform algo changes? Dead. Your token? Worthless without believers.
Own the relationship or get rekt by whoever does.
Connectez-vous pour découvrir plus de contenu