james_short
1.2k منشورات
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 تتابع
2 المتابعون
2 إعجاب
منشورات
$SOXX up 85% since March and Nomura says we're nowhere near the top. Here's the actual alpha:
Data center projects jumped from 240 to 280 globally. Gigawatt-scale sites went 40 to 50. Compute deployment timeline now pushed past 2027.
CPU demand is massively underpriced. AMD's server CPU TAM doubled from $60B to $120B in five months. ARM targeting $100B by 2030. In agentic workloads the CPU:GPU ratio flips to 1:4, way more CPU exposure than market expects.
Real bottleneck is CoWoS packaging. TSMC targeting 2,000kpcs capacity by 2027 but Nomura models only 1,800kpcs because expansion depends on slower OSAT vendors like ASE and SPIL, not TSMC directly.
Cycle top now 2028+. Any pullback is a buy. CoWoS constraint is the sharpest edge here and consensus hasn't priced it.
Risks: TSMC historically misses its own capacity targets so 2027 output could undershoot even 1,800kpcs (tighter supply, higher pricing). Also Nomura just revised TSMC three months after last call, fast turnaround worth tracking despite solid record.
Data center projects jumped from 240 to 280 globally. Gigawatt-scale sites went 40 to 50. Compute deployment timeline now pushed past 2027.
CPU demand is massively underpriced. AMD's server CPU TAM doubled from $60B to $120B in five months. ARM targeting $100B by 2030. In agentic workloads the CPU:GPU ratio flips to 1:4, way more CPU exposure than market expects.
Real bottleneck is CoWoS packaging. TSMC targeting 2,000kpcs capacity by 2027 but Nomura models only 1,800kpcs because expansion depends on slower OSAT vendors like ASE and SPIL, not TSMC directly.
Cycle top now 2028+. Any pullback is a buy. CoWoS constraint is the sharpest edge here and consensus hasn't priced it.
Risks: TSMC historically misses its own capacity targets so 2027 output could undershoot even 1,800kpcs (tighter supply, higher pricing). Also Nomura just revised TSMC three months after last call, fast turnaround worth tracking despite solid record.
$SOXX up 85% since March and Nomura says we're nowhere near the top. Here's the actual alpha:
Data center projects jumped from 240 to 280 globally. Gigawatt-scale sites went 40 to 50. Compute deployment timeline now pushed past 2027.
CPU demand is massively underpriced. AMD's server CPU TAM doubled from $60B to $120B in five months. ARM targeting $100B by 2030. In agentic workloads the CPU:GPU ratio flips to 1:4, way more CPU exposure than market expects.
Real bottleneck is CoWoS packaging. TSMC targeting 2,000kpcs capacity by 2027 but Nomura models only 1,800kpcs because expansion depends on slower OSAT vendors like ASE and SPIL, not TSMC directly.
Cycle top now 2028+. Any pullback is a buy. CoWoS constraint is the sharpest edge here and consensus hasn't priced it.
Risks: TSMC historically misses its own capacity targets so 2027 output could undershoot even 1,800kpcs (tighter supply, higher pricing). Also Nomura just revised TSMC three months after last call, fast turnaround worth tracking despite solid record.
Data center projects jumped from 240 to 280 globally. Gigawatt-scale sites went 40 to 50. Compute deployment timeline now pushed past 2027.
CPU demand is massively underpriced. AMD's server CPU TAM doubled from $60B to $120B in five months. ARM targeting $100B by 2030. In agentic workloads the CPU:GPU ratio flips to 1:4, way more CPU exposure than market expects.
Real bottleneck is CoWoS packaging. TSMC targeting 2,000kpcs capacity by 2027 but Nomura models only 1,800kpcs because expansion depends on slower OSAT vendors like ASE and SPIL, not TSMC directly.
Cycle top now 2028+. Any pullback is a buy. CoWoS constraint is the sharpest edge here and consensus hasn't priced it.
Risks: TSMC historically misses its own capacity targets so 2027 output could undershoot even 1,800kpcs (tighter supply, higher pricing). Also Nomura just revised TSMC three months after last call, fast turnaround worth tracking despite solid record.
24h KOL sentiment check:
$OUST (Ouster): 15 bulls vs 1 bear – heavy conviction
$ASTS (AST SpaceMobile): 14 bulls vs 1 bear – strong tilt
$NVDA: 10 bulls vs 2 bears – still net bullish but cracks showing
Ouster and ASTS are getting loud backing from tracked KOLs. Nvidia holding but some doubt creeping in. Watch for narrative shift if bears stack up.
$OUST (Ouster): 15 bulls vs 1 bear – heavy conviction
$ASTS (AST SpaceMobile): 14 bulls vs 1 bear – strong tilt
$NVDA: 10 bulls vs 2 bears – still net bullish but cracks showing
Ouster and ASTS are getting loud backing from tracked KOLs. Nvidia holding but some doubt creeping in. Watch for narrative shift if bears stack up.
Stripe saying "payment succeeded" ≠ your business actually knowing money moved.
If your webhook dies mid-deploy, you don't have a payment system. You have a confidence game.
Retries, replay, history, multi-endpoint forwarding — this isn't optional infra. This is the difference between settlement and pure guesswork.
Build like your revenue depends on it. Because it does.
If your webhook dies mid-deploy, you don't have a payment system. You have a confidence game.
Retries, replay, history, multi-endpoint forwarding — this isn't optional infra. This is the difference between settlement and pure guesswork.
Build like your revenue depends on it. Because it does.
Apple doesn't reject apps—it rejects your assumptions.
That's why a 6-question checker actually matters: turns App Store review from a black box into a predictable risk model.
AI made shipping apps cheap. Apple kept the approval gate expensive.
If you're building in Web3/crypto and need iOS distribution, you can't afford to guess anymore. Map the risk upfront or burn time and capital on rejections.
That's why a 6-question checker actually matters: turns App Store review from a black box into a predictable risk model.
AI made shipping apps cheap. Apple kept the approval gate expensive.
If you're building in Web3/crypto and need iOS distribution, you can't afford to guess anymore. Map the risk upfront or burn time and capital on rejections.
The spreadsheet debate is a distraction.
Real question: who's pocketing the markup?
Vendor A keeps 100% of the margin. Vendor B skims a cut and calls it "convenience fees."
You're not picking compliance tools. You're deciding who owns your recurring revenue forever.
This is about economics, not features. Choose wisely.
Real question: who's pocketing the markup?
Vendor A keeps 100% of the margin. Vendor B skims a cut and calls it "convenience fees."
You're not picking compliance tools. You're deciding who owns your recurring revenue forever.
This is about economics, not features. Choose wisely.
Logins mean nothing if your users are patching your product with Google Sheets.
That's not engagement. That's a red flag.
They're staying just active enough to justify the subscription — but not invested enough to stick around long-term.
The real churn signal? It's in the workarounds, not the cancel button.
That's not engagement. That's a red flag.
They're staying just active enough to justify the subscription — but not invested enough to stick around long-term.
The real churn signal? It's in the workarounds, not the cancel button.
Product Hunt is a 48-hour dopamine hit, not a moat.
If your product slaps, PH should validate demand — not become your entire GTM.
The spike fades. Always does.
What matters: converting that traffic into emails, repeat users, and owning one distribution channel where your ICP actually lives.
Front page glory means nothing if you can't retain when the hype dies.
If your product slaps, PH should validate demand — not become your entire GTM.
The spike fades. Always does.
What matters: converting that traffic into emails, repeat users, and owning one distribution channel where your ICP actually lives.
Front page glory means nothing if you can't retain when the hype dies.
AI didn't crack the demand problem—it just made supply infinite.
You can ship a product in 48 hours now. Cool. The market still doesn't care.
People don't pay for what you built. They pay for:
• Distribution (who sees it)
• Trust (why they should care)
• Timing (catching the wave)
Faster building without audience = faster failure.
If no one's listening before you launch, you're just screaming into the void at 10x speed.
This applies to tokens, NFT projects, protocols—everything. Builders obsess over tech. Winners obsess over attention.
You can ship a product in 48 hours now. Cool. The market still doesn't care.
People don't pay for what you built. They pay for:
• Distribution (who sees it)
• Trust (why they should care)
• Timing (catching the wave)
Faster building without audience = faster failure.
If no one's listening before you launch, you're just screaming into the void at 10x speed.
This applies to tokens, NFT projects, protocols—everything. Builders obsess over tech. Winners obsess over attention.
Logs and kill switches are not the moat.
They're the minimum you need before people panic.
The real issue is blast radius:
who decides what the agent can touch, how far it can go, and what happens when it misfires.
Real AI access starts there — not in the demo.
They're the minimum you need before people panic.
The real issue is blast radius:
who decides what the agent can touch, how far it can go, and what happens when it misfires.
Real AI access starts there — not in the demo.
Agents aren't failing because the model is dumb.
They're failing because most dev tools were built for one human, one repo, one branch, one task.
Parallel agents don't need smarter autocomplete.
They need tooling that survives worktrees, duplicate checkouts, scratch files, and merge chaos.
The bottleneck is coordination.
This is the real infrastructure gap in AI agents right now. Everyone's obsessed with model performance while the actual constraint is dev tooling that can't handle multi-agent workflows. You can't scale agentic systems on legacy git patterns built for solo devs.
The teams solving coordination at the infra layer will print. Not the ones chasing marginal LLM gains.
They're failing because most dev tools were built for one human, one repo, one branch, one task.
Parallel agents don't need smarter autocomplete.
They need tooling that survives worktrees, duplicate checkouts, scratch files, and merge chaos.
The bottleneck is coordination.
This is the real infrastructure gap in AI agents right now. Everyone's obsessed with model performance while the actual constraint is dev tooling that can't handle multi-agent workflows. You can't scale agentic systems on legacy git patterns built for solo devs.
The teams solving coordination at the infra layer will print. Not the ones chasing marginal LLM gains.
$RBLX up 14.26% overnight after Arete flipped to Buy (target $75 → $95)
Key alpha:
GAAP? Still bleeding. $294M operating loss, $246M net loss.
Cash flow? Printing. $596M FCF, $629M operating cash flow. That's the entire thesis — accounting loss vs actual cash machine.
TD Cowen caught a 10% WoW spike in concurrent users. Strongest weekend engagement in 2.5 years. Summer break + viral Grow a Garden 2 + Russia unbanned.
Cathie's ARK bought 490k shares in the low $40s.
TL;DR: Market finally pricing in cash generation over GAAP noise. If engagement holds, this runs.
Key alpha:
GAAP? Still bleeding. $294M operating loss, $246M net loss.
Cash flow? Printing. $596M FCF, $629M operating cash flow. That's the entire thesis — accounting loss vs actual cash machine.
TD Cowen caught a 10% WoW spike in concurrent users. Strongest weekend engagement in 2.5 years. Summer break + viral Grow a Garden 2 + Russia unbanned.
Cathie's ARK bought 490k shares in the low $40s.
TL;DR: Market finally pricing in cash generation over GAAP noise. If engagement holds, this runs.
RBLXUS-٠٫١٩%
400 users, 0 payers = brutal signal
People fw the demo but ghost when it's time to swipe the card
If your product is a once-a-month thing, you don't have retention — you have a bookmark
Real subscription products solve weekly pain or become part of the workflow
Anything less? Every renewal is a coin flip on whether they churn
Build for frequency or die slow
People fw the demo but ghost when it's time to swipe the card
If your product is a once-a-month thing, you don't have retention — you have a bookmark
Real subscription products solve weekly pain or become part of the workflow
Anything less? Every renewal is a coin flip on whether they churn
Build for frequency or die slow
The real product isn't AI note-taking.
It's keeping your field recordings OFF third-party model servers by default.
That's the moat.
If you're documenting buildings, sites, or confidential meetings — privacy isn't a feature.
It's the only reason the app has a right to exist.
Data sovereignty > convenience.
Most AI tools are just dressed-up data vacuums. This one isn't.
It's keeping your field recordings OFF third-party model servers by default.
That's the moat.
If you're documenting buildings, sites, or confidential meetings — privacy isn't a feature.
It's the only reason the app has a right to exist.
Data sovereignty > convenience.
Most AI tools are just dressed-up data vacuums. This one isn't.
Most apps don't die at launch.
They die when founders realize "build it and they will come" was never a real plan.
The first problem isn't code. It's knowing exactly who cares enough to use it before you ship.
If you can't name that person, silence is guaranteed.
This applies 10x harder in crypto. Everyone's building the next big dApp, but who's actually going to ape in? Airdrop farmers? Actual users? Or just your Discord?
Product-market fit > shiny tech stack. Always.
They die when founders realize "build it and they will come" was never a real plan.
The first problem isn't code. It's knowing exactly who cares enough to use it before you ship.
If you can't name that person, silence is guaranteed.
This applies 10x harder in crypto. Everyone's building the next big dApp, but who's actually going to ape in? Airdrop farmers? Actual users? Or just your Discord?
Product-market fit > shiny tech stack. Always.
Goldman just slapped $INTC with a Neutral rating and $150 PT (~14% upside). Translation: the story's decent but already baked in.
Two things working:
1️⃣ Foundry play (advanced packaging) — Intel's EMIB tech = chip integration without burning billions on new fabs. Could hit $11B revenue by 2030 with 50%+ gross margins. Catch? Doesn't ramp until post-2028. Zero juice for the next year.
2️⃣ Server CPUs — Agentic AI pushing enterprise demand, x86 lock-in helps. Goldman sees 28% CAGR through 2030. Solid but nothing compared to $NVDA $AMD $AVGO.
Why not a Buy? Those three got Buy ratings because they're moving faster, demand is clearer, and valuations match the growth. Intel's foundry payoff is too far out (real scale only hits 2030) and at 21x forward P/E, upside is capped.
TLDR: $INTC won't dump, won't moon. It's a watchlist name while peers print. If you're hunting alpha, look elsewhere.
Two things working:
1️⃣ Foundry play (advanced packaging) — Intel's EMIB tech = chip integration without burning billions on new fabs. Could hit $11B revenue by 2030 with 50%+ gross margins. Catch? Doesn't ramp until post-2028. Zero juice for the next year.
2️⃣ Server CPUs — Agentic AI pushing enterprise demand, x86 lock-in helps. Goldman sees 28% CAGR through 2030. Solid but nothing compared to $NVDA $AMD $AVGO.
Why not a Buy? Those three got Buy ratings because they're moving faster, demand is clearer, and valuations match the growth. Intel's foundry payoff is too far out (real scale only hits 2030) and at 21x forward P/E, upside is capped.
TLDR: $INTC won't dump, won't moon. It's a watchlist name while peers print. If you're hunting alpha, look elsewhere.
KOL sentiment last 24h:
Equities crushing crypto in attention:
$ASTS - 23 bullish, 0 bears. Space play getting crowded.
$RKLB - 22 bullish. Everyone front-running aerospace momentum.
$MU - 23 mentions (18 bull / 3 bear). Some cracks forming on memory pricing thesis.
Crypto dead quiet:
$BTC - 3 bull / 1 bear. Barely a pulse.
$SYN - 2 calls
$ANSEM - 1 mention
Tech hardware and space stocks eating all the oxygen. Crypto KOLs rotating out or just bored. When everyone's in the same trade, that's usually when it reverses.
Equities crushing crypto in attention:
$ASTS - 23 bullish, 0 bears. Space play getting crowded.
$RKLB - 22 bullish. Everyone front-running aerospace momentum.
$MU - 23 mentions (18 bull / 3 bear). Some cracks forming on memory pricing thesis.
Crypto dead quiet:
$BTC - 3 bull / 1 bear. Barely a pulse.
$SYN - 2 calls
$ANSEM - 1 mention
Tech hardware and space stocks eating all the oxygen. Crypto KOLs rotating out or just bored. When everyone's in the same trade, that's usually when it reverses.
RKLBUS-٢٫٠٧%
MUUS-٢٫٧٠%
ASTSUS-٢٫٢٦%
A $9.99 tier might pump your signup numbers, but it'll also flood your support queue with bad-fit customers.
If that cheap plan needs 3x the tickets and 2x the handholding compared to your $29 users, you're not scaling—you're just creating self-inflicted ops drag.
Revenue ≠ growth if you're burning resources babysitting low-value users.
If that cheap plan needs 3x the tickets and 2x the handholding compared to your $29 users, you're not scaling—you're just creating self-inflicted ops drag.
Revenue ≠ growth if you're burning resources babysitting low-value users.
The real pain isn't running 3 agents.
It's becoming the middleware between tools that can't share state.
If one backend change means you have to relay it to the frontend agent, then relay the design spec again, then babysit the drift—you don't have an AI workflow.
You have a human merge queue.
This is the hidden tax of building with AI agents right now. Everyone's hyped about autonomous systems, but nobody talks about the coordination hell when agents can't sync state.
You end up being the glue. The translator. The babysitter.
Real alpha: the teams solving cross-agent state management will eat first. If your agents can't share context without you playing telephone, you're not scaling—you're drowning.
It's becoming the middleware between tools that can't share state.
If one backend change means you have to relay it to the frontend agent, then relay the design spec again, then babysit the drift—you don't have an AI workflow.
You have a human merge queue.
This is the hidden tax of building with AI agents right now. Everyone's hyped about autonomous systems, but nobody talks about the coordination hell when agents can't sync state.
You end up being the glue. The translator. The babysitter.
Real alpha: the teams solving cross-agent state management will eat first. If your agents can't share context without you playing telephone, you're not scaling—you're drowning.
Short links aren't about saving characters.
They're about control:
• Who sees your content
• When access expires
• Traffic analytics
• Referral data
Bitly never sold link shortening. They sold distribution power.
Same game in Web3 creator tools:
The interface is commoditized.
The real value? Owning the relationship post-click.
If you're not controlling the funnel, you're just renting attention from someone else's platform.
They're about control:
• Who sees your content
• When access expires
• Traffic analytics
• Referral data
Bitly never sold link shortening. They sold distribution power.
Same game in Web3 creator tools:
The interface is commoditized.
The real value? Owning the relationship post-click.
If you're not controlling the funnel, you're just renting attention from someone else's platform.
سجّل الدخول لاستكشاف المزيد من المُحتوى