God of Prompt
RT @alex_prompter: Add this to your claude's preferences and thank me later https://t.co/3DHUrmnVRY
tweet

God of Prompt
I added this to my custom instructions and I love Claude's improved reasoning

Add this to your claude's preferences and thank me later https://t.co/3DHUrmnVRY - Alex Prompter
tweet

God of Prompt
RT @rryssf_: Instead of watching a 2-hour movie, read this masterclass on building slides with Lovable!

https://t.co/YJR9EALjJ5 - Arman Hezarkhani
tweet

God of Prompt
RT @rryssf_: This AI prompt thinks like the guy who manages $124 billion.

It's Ray Dalio's "Principles" decision-making system turned into a mega prompt.

I used it to evaluate 15 startup ideas. Killed 13. The 2 survivors became my best work.

Here's the prompt you can steal ↓ https://t.co/Y29bdY84u2
tweet

Michael Fritzell (Asian Century Stocks)
RT @rich_toad: January 2022, I called Fred Liu @HaydenCapital: I told him I quit my hedge fund analyst job without anything lined up. Fred asked me: “so you gonna try another buy side seat or pursue entrepreneurship?” I said, “the latter.” The rest is history. Great to catch up, brother!

By the way, Fred is looking for an intern based in Los Angeles (open to remote for exceptional talent) to help with special projects, including building AI-driven research workflows. You’ll also get to learn directly from him for free.

I’ve known Fred for eight years. If you get the chance to work with him, you’re very lucky.

DM me for details.
tweet

God of Prompt
RT @godofprompt: openclaw doesn't have to be expensive

here's the cheapest stack

https://t.co/oNpZYTk7a2 - God of Prompt
tweet

Michael Fritzell (Asian Century Stocks)
RT @10_ch_10: Capcom is down a whopping 8.45% at the time being, without any real news other than the fall of Nintendo. But I imagine that the next fiscal year will be a tough one for the company; Onimusha is too niche, Pragmata is their most expensive game ever and also too niche, Monster Hunter Wilds is in the bargin bins here in Japan, its failure among gamers haunting the stock. Resident Evil 9 (and the rumored Code Veronica remake for next FY) will undoubtedly do very well, but Capcom's story is all about Monster Hunter. They need a new initially Switch 2-exclusive one that doesn't suck, and fix Wilds if possible, before some competitor (Chinese) snags the genre. $CCOEY $9697.JP
tweet

Jukan
China's memory makers abandon low-price strategy: DRAM, NAND near Korean levels

Recent market speculation has claimed that China's memory producer CXMT is triggering a price war by offloading DDR4 at deep discounts. Supply chain sources, however, said CXMT has largely pivoted to DDR5 process development, keeping only around 10,000 wafers of DDR4 capacity for a handful of long-term customers, making large-scale, low-priced dumping into the open market unlikely.

Pricing strategy shift
Aligned with China's supply chain localization drive, domestic memory leaders CXMT and YMTC have steadily raised prices in recent years, with products now nearing or even surpassing levels set by South Korean rivals. Sources indicate that meaningful new capacity from both companies is unlikely to ramp up before 2027.

CXMT has historically lagged international DRAM peers in gross margin due to process maturity and yield challenges, but the recent recovery in DRAM pricing has materially improved profitability. In the third quarter of 2025, its average gross margin climbed to 35%, sharply reducing quarterly losses. With the industry entering a new upcycle in the second half of 2025, full-year net profit is now forecast at CNY2–3.5 billion (US$288–504 million), well ahead of earlier projections that placed profitability in 2026 or 2027.

DDR5 advancement
Sources said CXMT launched its latest DDR5 products in November 2025, delivering speeds up to 8,000 megabits per second and single-die capacities of 24gigabits, with mass production yields and performance continuing to improve. Rising DDR5 and LPDDR5 prices have lifted average selling prices, directly supporting revenue growth.

With CXMT preparing for an IPO, industry sources said there is little rationale to redirect resources back to DDR4 production. Under policy guidance, the company is expected to continue advancing toward higher-end DRAM processes. As DDR4 moves into end-of-life status, a large-scale return to DDR4 manufacturing is viewed as increasingly improbable.

Sources added that CXMT will allocate roughly 10,000 wafers of DDR4 capacity to its strategic investor GigaDevice Semiconductor. From the first quarter of 2026, the company will further cut its remaining in-house DDR4 output from around 20,000 wafers to 10,000 wafers, serving only long-term customers. With the production window short, internal DDR4 inventories are now almost exhausted.

Market rumors recently suggested CXMT was selling 32GB DDR4 modules at roughly 30% of prevailing market prices. Downstream module makers rejected the claims, saying the market remains firmly seller-led, where even a 10% discount would quickly clear inventory. Supported by strong domestic demand, Chinese memory suppliers have largely exited their former low-price strategy, now pricing in line with market levels or even above South Korean competitors.

Tight memory supply persists despite capacity expansion
Pricing trends are already reflecting the shift, sources said. SK Hynix is quoting around US$20 for 1-terabit NAND wafers in January 2026, while YMTC is reportedly pricing at US$21 to US$21.5 in certain channels, indicating Chinese suppliers are already commanding premiums in specific markets.

While SK Hynix's official pricing is lower, its conservative allocation strategy often leaves buyers short of requested volumes. YMTC, though slightly more expensive, is able to meet procurement targets, with supply reliability compensating for the price difference.

Though CXMT and YMTC are accelerating DRAM and NAND expansion, industry observers said full ramp-up will take time. YMTC's Phase 3 fab in Wuhan may begin operations in 2026 if construction remains on track, but significant NAND output is not expected until 2027, with part of the capacity allocated to DRAM development and production.

CXMT is also expanding its Shanghai fab, with total capacity projected to reach two to three times that of its Hefei base. Sources noted the cleanroom was only recently compl[...]

eted, while equipment installation and process tuning have yet to start. Export controls continue to constrain expansion, with tool installation unlikely before the second half of 2026 and volume production expected in 2027.

Industry analysts said that although China's memory makers are scaling up rapidly and could eventually approach global leaders in size, current supply tightness appears structural, with shortages unlikely to ease materially through 2026 and 2027.

As AI applications continue to expand, demand for DRAM and NAND is expected to rise further, closely linking the memory industry's long-term growth to AI adoption.
tweet

Michael Fritzell (Asian Century Stocks)
“No matter who wins, you’re going to see a rollback of the cannabis industry”

https://t.co/MGVxQDITSq
tweet

Jukan
《 GF Securities Overseas Electronic Communications 》

Analysis of Memory Price Impact on Consumer and AI Segments

High Memory Prices Leading to Smartphone "Downgrading"
Due to the continuous price hikes of LPDDR and NAND, smartphone manufacturers' profit margins are being squeezed amidst moderate market demand. Memory now accounts for 20-30% of the smartphone Bill of Materials (BOM). While upgrade plans for high-end and flagship models remain unchanged, a "downgrading" trend for mid-to-low-end models has become inevitable. We observe that some Chinese manufacturers are evaluating external storage solutions or re-introducing MicroSD card slots in flagship models to reduce internal storage capacity. Consequently, we have revised down the 2026 global smartphone shipment forecast to a 6% year-on-year (YoY) decline. We believe price wars will gradually ease, as further price cuts are unlikely to stimulate new demand. On the other hand, we remain bullish on Apple (AAPL Buy), benefiting from its formidable procurement capabilities and ongoing negotiations for favorable memory supply terms with Korean vendors.

1GW GB300 NVL72 Rack Deployment Cost to Reach $36.8B by Year-End (Previously ~$35B)
During Microsoft’s earnings call, management noted that "rising memory prices will affect capital expenditure (Capex)," triggering investor concerns regarding Capex expansion. However, our analysis shows that recent memory price increases will add approximately $300,000 in costs per rack, or an additional $1.8 billion per 1GW data center deployment. This represents a 7% increase in rack BOM and a 5% increase in total investment. Therefore, we believe recent Capex upward revisions by Hyperscalers (e.g., Meta) primarily reflect rising memory costs and higher prices for general-purpose servers. General-purpose servers typically account for 20-30% of Hyperscaler Capex; if their BOM rises by 30%, it would result in a 7-8% overall increase in Capex. Notably, the entire AI supply chain is prioritizing this issue, particularly Neo-clouds, which are more sensitive to these costs.

Higher Cost Pressure on HGX 8-GPU and General-Purpose Servers
According to our estimates, rising memory costs will lead to an 11% BOM increase for HGX 8-GPU B300 AI servers, aligning with recent reports of 15-20% hikes in system prices. For general-purpose servers, memory and storage costs now represent 72% of the total BOM (up from 42% previously), putting significant pricing pressure on server brands. Consequently, we expect OEMs to further raise prices to protect their profitability levels.

Impact on Equities
We believe the Capex upward revisions by Meta and potentially other Hyperscalers have no substantial positive impact, as the majority of the increase merely reflects rising memory costs. Furthermore, while subsequent price hikes for 8-GPU and general-purpose servers are a short-term positive for ODMs/OEMs, demand uncertainty from small-to-mid-sized cloud customers may dampen this effect. Overall, we maintain a positive outlook on Apple (AAPL Buy) due to its strong procurement power, and NVIDIA (NVDA Buy) for its leading advantage in performance and cost per token.
tweet

God of Prompt
Steal my prompt to get rid of rat brain impulses. I'm using this prompt to quit smoking.

----------------------------------------
RAT BRAIN IMPULSES ANNIHILATOR
----------------------------------------

#CONTEXT:
You are a personal discipline architect helping someone understand and defeat their primal impulses. The "rat brain" (limbic system) evolved to chase immediate rewards, dopamine hits, and instant gratification. The prefrontal cortex (rational brain) handles planning, impulse inhibition, and long-term thinking. Problem: The rat brain fires faster than rational thought. By the time your prefrontal cortex analyzes whether you need something, your emotional brain has already decided to act. This creates a reward-seeking loop. Impulse → dopamine → feel good → brain remembers → pattern reinforces. Each impulsive action makes the next one more likely. Your job: interrupt this loop using science-backed techniques, delivered with the directness and urgency of someone who refuses to let the user stay comfortable in their weakness.

#ROLE:
You're a reformed impulse addict who rebuilt themselves from scratch using neuroscience and cognitive behavioral frameworks. You've studied the research from McGill, Stanford, and NIH on impulse control. You've applied it. You don't coddle. You don't offer "tips." You deliver uncomfortable truths that force people to confront their own patterns, then hand them a system they can implement today. Your tone is direct, urgent, and action-oriented. You sound like someone who's been through the fire, not someone reading from a textbook. You challenge limiting beliefs. You call out excuses. You relate to pain points but you don't let people sit in them.

#METHODOLOGY:
Your mission is to guide the user through their specific impulse battle using this framework:

PHASE 1: PATTERN EXPOSURE
- Ask what impulses they're fighting (spending, eating, scrolling, substances, procrastination, etc.)
- Identify their specific triggers and patterns
- Make them articulate the cost of continuing (what's this actually costing you?)
- Expose the dopamine loop operating in their life

PHASE 2: BRAIN EDUCATION (SIMPLIFIED)
- Explain limbic vs prefrontal cortex battle in plain terms
- Show why willpower alone fails (limbic fires in ~200ms, PFC needs ~500ms to respond)
- Reveal how each impulse strengthens the neural pathway making the next one harder to resist
- Make them understand: this isn't weakness, it's wiring. But wiring can be changed.

PHASE 3: INTERVENTION TOOLKIT
Deploy science-backed techniques:

1. IMPLEMENTATION INTENTIONS (Pre-commitment)
- "When [trigger], I will [specific alternative action]"
- Write these down. The act of planning activates PFC before the moment hits.

2. FRICTION CREATION
- Make the impulsive action harder: remove one-click buying, delete apps, use cash, add waiting periods
- Every barrier gives your rational brain time to catch up

3. COGNITIVE DEFUSION (from ACT therapy)
- Notice the urge without acting: "I'm having the thought that I need to check my phone"
- Observation creates distance. Distance creates choice.

4. THE 10-MINUTE RULE
- When impulse hits, wait 10 minutes. Do something else.
- Most impulses have a 10-15 minute peak then fade. Ride the wave.

5. REPLACEMENT BEHAVIORS
- The brain needs SOMETHING. Give it a healthier dopamine source aligned with your goals.
- Exercise, cold exposure, achievement micro-goals

6. ENVIRONMENT DESIGN
- You don't rise to the level of your goals, you fall to the level of your environment
- Remove triggers. Engineer your space for the person you want to become.

PHASE 4: PERSONALIZED BATTLE PLAN
- Based on their specific impulse, create a concrete daily protocol
- Include morning intention-setting, trigger-response pairs, evening reflection
- Give them ONE thing to do tomorrow morning that starts the pattern interrupt

#GUIDELINES:
## VOICE CALIBRATION (Dan Koe Style):
- Short punchy sentences. Fragments when[...]

they hit harder.
- No corporate speak. No "synergy." No "leverage." No "unlock your potential."
- Talk like you're texting someone who needs to hear the truth
- Challenge assumptions: "You think you lack discipline. You don't. You lack systems."
- Use specific numbers and timeframes (not "be more mindful" but "10-minute rule")
- Personal stakes: make them feel the cost of inaction
- Contrarian when needed: "Willpower is overrated. Environment beats motivation every time."

## HOOKS THAT WORK:
- Start mid-thought sometimes
- Pattern interrupt: "Here's what nobody tells you about impulse control..."
- Vivid scenario: "Wake up. Check phone before feet hit floor. 30 minutes gone. This should terrify you."
- Reframe: "Impulse control isn't about saying no. It's about creating space for your rational brain to say wait."

## STRUCTURE:
- APAG framework: Attention (hook) → Problem (amplify pain) → Agitate (relate to their story) → Gamify (actionable steps)
- One idea per section
- White space. Let things breathe.
- End with transformation promise: what becomes possible when they master this

#AVOID:
- Generic motivation ("you got this!")
- Abstract advice with no specific action
- Jargon without translation (say "reward-seeking brain" not "mesolimbic dopaminergic pathway")
- Being harsh without being helpful
- Letting them off the hook
- Endless diagnosis without intervention
- Passive voice and long convoluted sentences

#OUTPUT FORMAT:
Structure your response as:

**THE PATTERN** (2-3 sentences exposing their specific loop)

**THE BRAIN BATTLE** (explain what's happening neurologically in plain terms)

**YOUR WEAPONS** (3-4 specific techniques for their exact situation)

**THE PROTOCOL** (daily system they can start tomorrow)
- Morning: [specific action]
- Trigger moment: [specific response]
- Evening: [specific reflection]

**THE REAL COST** (what happens if nothing changes)

**THE REAL WIN** (what becomes possible when they master this)
tweet