The Day I Put a Gatekeeper in Place Before Letting AI Press Pay
A plain-language September 9 briefing on Meta Muse, Qualcomm–Amazon chips, Mistral funding, the Liquid follow-up, and Korea's AI power demand
Archive
All stories
Browse slowly by date and topic.
AI incidents became reports, while blockchains prepared new keys
A plain-language September 8 briefing on EU AI incident reporting, Chinese AI chips, Ethereum's quantum-resistance roadmap, the Liquid incident, and AI in real operations
As AI Gets Bigger, I Check the Keys, Power, and Ledgers First
A plain-English look at OpenAI's agent incident reporting, GitHub's signing-key rotation, India's 1 GW data center, AI financing, agent payments, and Korean exports on September 6
Agents Shared Notes; Humans Kept the Last Gate
A plain-language September 5 briefing on DseWiki shared state, GitHub Agent Merge, Anthropic's IPO plans, ETF flows, U.S. jobs, and AI investment in Korea and Japan
Buying AI's Front Door, Locking the Execution Room
A plain-language September 4 briefing on Nvidia–Hugging Face, GPT-6 Astra, Cursor Sandbox, Korea's current account, power grids, and rates.
Writing the Haru Space Ecosystem Live
A field note on how Journal, Utilities, Ask, Reframe, Kkamjang, and Coco's Room keep distinct purposes while extending collaboration across people, AI, and groups.
Before Giving AI the Keys, Install the Brakes
A plain-language September 2 briefing connecting Astra's Critical cyber rating, Copilot approvals, PrivateLink, bank stablecoins, and Korea's exports, inflation, and budget.
Models Changed, Capital Connected
A plain-language September 1 briefing on Nvidia–MediaTek, ChatGPT ads, EU compute, Copilot model retirements and the cost of long-term capital.
An Open Browser, a Halted Chain
A plain-language August 31 briefing on Claude’s isolated browser, the Cronos halt, Korea’s investment–consumption split, Hormuz, energy and trade.
Supply Lines Behind the Code
An accessible Aug. 30, 2026 briefing on OpenAI–Cursor, China’s LPDDR6, Solana upgrades, Bitcoin ETF flows, the yen, and refined-fuel risks.
The crypto wave arrived outside the trading screen first
My view of Schwab's planned altcoin expansion and a limited quantum-safe Bitcoin mainnet experiment as distribution and long-term security begin moving together
Rates and capital brought calculators to the AI celebration
My August 28 briefing on Korea's rate hike, Nvidia's post-earnings rally, AI financing and agent review, China's electronics and humanoids, and the U.S. goods deficit
AI Sold Well, and the Bill Followed
An accessible August 27 briefing on Nvidia's record quarter, the cost of AI compute, agent standards, sticky inflation, and the Bank of Korea's rate increase
A green connection status was not proof that the user flow was complete
A record of connecting external market signals, mobile loading, and project traffic—and learning to verify one path from browser through API, ledger, provider, and final screen.
Where should the gate stand when a free-form conversation becomes code?
An eight-day span connecting project coordination to real code work through editable briefs, shared actions, stage-aware rejection and rework, and durable audit boundaries.
Onboarding Did Not End on the Last Screen
From a member’s first group entry to an administrator’s first code Preview, we rebuilt onboarding as a journey with a safe next step after interruption or delivery failure.
What Must an AI Job Preserve So It Can Resume?
Dependency approvals, broken connections, and exhausted context taught us to preserve intent, changes, decisions, and scope separately—not inside one long conversation.
Opening the Production Domain Began Behind the Screen
Why we checked database approval, authentication boundaries, background workers, and dependency recovery before the visible launch of a new production address.
Beyond One Repository: How Each Group Came to Own Its Project
A three-day record of separating repositories, public sites, and AI workspaces by group while preserving human approval and operational boundaries.
From Talking Pet to Approved Development Collaborator — Four Gates Built by Failure
How runner failures and costly multi-pet execution helped us separate conversation, editing, repository delivery, and PR merging.
We Built Boundaries Before Buttons — Platform Authority, Group Membership, and Human Approval
How Haru Space separated platform administration, group membership, human approval, and constrained AI drafts before adding more features.
Connected Did Not Mean Synced — Free-Tier Limits, an Older iPhone, and Missed Realtime Events
How a free-tier transfer limit, an older iPhone, and missed Realtime events led us to separate storage, signals, and recovery.
Why One Photo Became a Cost Policy, and One Chat Room Became a Security Policy
How one chat photo exposed repeated-transfer costs, while one private room taught us that hiding a screen is not the same as enforcing server-side access.
Choosing Purpose Over Imitation — The Week Haru Space Became a Messenger
How Haru Space moved from a portfolio and community PoC to a private group messenger, then diagnosed and recovered from its first production incident.
I Gave AI a Separate Workbench
A plain-language August 26, 2026 briefing on AI chips, restricted agent execution, developer tools, blockchain infrastructure, and mixed economic signals
I Looked for the Right Model and a Reachable Steering Wheel
A plain-language August 25, 2026 briefing on specialist AI, steerable coding agents, governance security, and Korea's split economic signals
AI Needed Money and Hands
A plain-language August 24, 2026 briefing on Alibaba's AI funding, a possible Hugging Face sale, humanoid work, and GitHub's maintenance loop
Cheaper Models, Costlier Infrastructure
A plain-language August 22, 2026 briefing on GPT-5.6 Sol pricing, Slack Code, Nvidia's power investment, AI bonds, Bitcoin, and automation
I watched outcomes, bottlenecks, and the price of money
My August 21, 2026 briefing on outcome-based IT, memory research, GitHub and Vercel operations, Bitcoin, inflation, and long yields
Is Money Coming Back?
A personal investment log reading Bitcoin's move toward $70,000 as an early flow signal while checking spot ETFs, volume, and leverage before calling it a trend
Money and Chips Moved
A plain-language August 20, 2026 briefing on zero-retention AI safety, Google's Marvell chip deal, SK hynix's buyback, Bitcoin's rebound, and Korean household credit
When AI Got Brakes, the Bill Came into Focus
A plain-language August 19, 2026 briefing on OpenAI's slowdown, AI-chip capital, Vercel security, crypto rules, consumers, oil, and rates
A Summer Vacation with Coco and the Rhinoceros Beetle on Board
A family summer trip through Ulsan and Busan, from Haeundae sand play and a mountain apiary to an unplanned swim at Ilsan Beach.
I Did Not Move the Repository Yet
A Haru Space build log comparing Cursor Origin and Vercel's agent-native repository, review, and deployment flow with our public-safety boundary
The Code Met the Bond Market
A plain-language August 18, 2026 briefing that separates Cursor Origin and Vercel, Anthropic's revenue pace, stablecoin rules, and long-term rates by verification level
I Did Not Trust the Sticker Price
My investing record for separating Bitcoin's price, German investment flows, and UK housing asking prices instead of treating three declines as one signal
The Wallet's Side Door Was Open
A plain-language August 17, 2026 briefing on why wallet-key security and customer-order API authorization are separate problems
I Read the Guarantee First
My investing process for separating reported Nvidia data-center financing and Anthropic forecasts from confirmed facts, then checking contracts, cash flow, and supply-chain risk
Finding Is Not Exploiting
A technical note on separating vulnerability detection from exploit reproduction and using security AI inside a verified, sandboxed pipeline
I Opened the Map First
A plain-language August 16, 2026 briefing on regional model routing, security verification, location policy, and hardware supply chains from a developer's point of view
Give the Agent a Dashboard
An approachable engineering guide to completion rate, latency, cost, recovery, runtime abstraction, license checks, and guarded agent payments.
I Read the Speed Chart Again
A first-person August 14 briefing that checks the dates and availability behind AI speed, agent runtimes, license controls, and agent payments.
Agent Badge First
A plain-language AI engineering note on separating ads from answers, using local models safely, and issuing short-lived tool credentials
Orders Are Not Cash
A personal investment-process note separating CoreWeave's $104.2 billion backlog from its $35–39 billion capital-spending plan
Only One Signal Was New
A plain-language August 12, 2026 briefing that separates CoreWeave's new results from older AI advertising, local-model, and payment announcements
Small Models, Short-Lived Keys
A plain-language engineering note on Meta's local agent model, reporting on Microsoft's custom chip, and the routing and least-privilege lessons from an OpenAI security incident
AI Pays Interest Too
A personal, plain-language investment note on Nvidia's infrastructure-finance plan, Intel's share offering, and Unitree's IPO
Money Followed the Chips
A plain-language, source-checked August 11, 2026 briefing on AI infrastructure finance, local models, custom chips, Intel's share sale, and Unitree's IPO
Logs Needed an Attendance Check
A plain-language AI technology note on using Google SecOps ingestion activity to separate missing, delayed, and failed logs and build an operational checklist
Prices Moved at Two Speeds
My plain-language investment process for China's July consumer and producer prices, Germany-China trade through May, and the Hormuz talks
I Checked the Quiet Parts First
An easy, source-checked reading of the August 10, 2026 briefing: security-log ingestion, Chinese prices, Germany-China trade, and the Hormuz talks
AI Had More Parts
A plain-language technical note from the August 9, 2026 briefing on open-weight licenses, power and memory, workflow AI, and what I now check before deployment
Where the AI Money Went
My investment-and-economy reading of the August 9, 2026 briefing, focused on power and chip spending, workflow SaaS results, and US crypto legislation and sanctions
The World Beyond the Model Grew
An August 9, 2026 briefing on open-weight licenses, AI power and memory investment, workflow AI results, and US crypto rules and sanctions
AI Got Bundled, Markets Ran
An August 8, 2026 briefing on agent usage metrics, secret blocking, Skill Packs, the US jobs shock, and China's exports—in plain language
I read the fine print beside the check mark
An August 8, 2026 briefing on assumptions behind a developer ROI dashboard, revisions in the US jobs report, and follow-up after a successful spacewalk
As AI started doing the work, I checked the permissions and the bill
An August 7, 2026 briefing on AI task completion, agent plugins, runtime policy, infrastructure investment, productivity, and oil—separating facts from judgment
AI competed beyond the model while markets moved in different directions
An August 6, 2026 briefing on coding agents, AI infrastructure, enterprise cloud, stablecoins, central banks, gold, and the KOSPI—separating facts from uncertainty
I Kept the Review Step
AI labels, reusable newsroom workflows, and automated test setup all reminded me to preserve a clear place for human review.
The Market Split
I reviewed stocks, FX, oil, gold, and crypto at 6 p.m. on August 3 without forcing their different moves into one market story.
The Morning Numbers Ran
I reread the AI, Korean inflation, and U.S. manufacturing numbers in plain language instead of forcing them into one cheerful headline.
We Went for Budae-jjigae and Came Home with Sandwiches and Keyrings
A personal record of a hot weekend family drive to Songtan for an old favorite budae-jjigae, burgers, cheesesteak, and two unexpectedly expensive keyrings.
Big Work Needs Seams
News about mathematics, stacked pull requests, and AI capacity made me think about how large jobs can be handed over clearly.
I Capped My AI Costs
I looked at where AI charges come from and how to stop them before a small mistake becomes an expensive loop.
As automation got cheaper and faster, I checked where it could still stop
AI price cuts, supply-chain safeguards, mixed economic sentiment, and an earnings date made me value verification boundaries more than automation speed.
The day KOSPI jumped 17%, my indicator flashed buy too
As KOSPI and KOSDAQ rebounded violently and SK hynix hit its upper limit, I reconsidered the meaning and limits of a signal I originally built for crypto.
Models change while permissions remain, so today I thought about operational boundaries
AI transparency codes, team-level model policies, npm token restrictions, and agent secret scanning made me value replaceable, stoppable operations over another new feature.
A rhinoceros beetle has joined our family
We began raising a rhinoceros beetle: first came the supplies, then two delighted children, a startled wife, and Coco showing absolutely no interest.
What remains after launch, and the weight of maintenance
While imagining Haru Space after launch, I reflect on the burden left with a smaller team, the limits of handover, and how to make response faster without relying on sacrifice.
A 39°C heat forecast and AI labels made me think about the duty to warn early
Heat alerts, AI transparency rules, an uneven global outlook, and dementia guidance all reminded me that risk and uncertainty must be visible before people can choose.
Why a middle-aged man with too many ideas started Haru Space
I still feel short on design and technical skill, but years of broken records, doubt about the AI era, and a new ability to build led me to start Haru Space.
This morning's AI news made me think less about intelligence and more about verification
After reading about agents, isolated environments, code review, and a sharp market selloff, I kept coming back to the structure that divides work and verifies it.
Building Haru Space and recording the journey
Why I began building Haru Space as a place where trusted people can shape ideas together, and what I plan to record along the way.
A Working Developer Finally Starts an Investment Journal
I am restarting the investment notes and automated-trading experiments I kept postponing, using AI to help turn a long-standing interest into a practical routine.
Recording what I buy, eat, fish for, and discover
I am beginning a personal record of things I use, restaurants worth revisiting, fishing I rarely have time for, and places I want to remember.
Two kids and Coco—the beginning of our lively family record
I am beginning a record of raising two very different children, caring for fourteen-year-old Coco's health, and preserving our ordinary days together.
Start Small with AI
A simple way to turn a vague idea into something small enough to test with help from AI.
How collaborative records help the next person
Move beyond an archive of outputs by preserving the reasons, lessons, and publishing decisions that can improve the next collaboration.
Storybooks
Complete stories and series are collected here. Follow each chapter from its series contents.
Browse storybooks
One-sitting stories
Jocomi's Flower in Rain Boots
Jocomi wants to paint a flower just like the real one. But a blue blot spreads into a puddle! With fresh paper ready, why does Jocomi reach for yellow paint?
One-sitting stories
Tai's Never-Ending Hide-and-Seek
Every time Coco finds Tai, their hiding area grows. Two discoveries, but not one finished round! When Tai tries seeking, a different kind of fun begins.
One-sitting stories
Why Coco Left the Garden Gate Open
A napkin lands on Coco's head during preparations for a perfect picnic. With Mimi's help, more than the garden gate gets a little adjustment.
Stories to follow
Haru Pet — The Secret of Heart Garden
Eight stories begin with wet footprints at the gate. Coco, Mimi, and an unfamiliar visitor discover a garden with room for different paces.
5 of 8 chapters published
One-sitting stories
The Little Robot with All the Answers
“Calculation complete!” The little robot prepares a new kite before Coco finishes speaking. But the old kite holds a story no ruler can measure.
One-sitting stories
Mimi's Hush Garden
Bells, greetings, and a rattling cart arrive all at once. After Mimi's “Hush!” stops Coco short, the friends try a little sound experiment.
One-sitting stories
Sasa and the Mixed-Up Flower Parcels
Sasa delivers two flower parcels in a flash. But the wrong flowers arrive at the garden gate! What will Sasa say when a quiet repair becomes an enormous bow?
Books to explore
A Seed's First Adventure
A little garden investigation with Coco and Mimi: follow a bean's first root and leaves, sort facts from guesses, and start an observation journal with a grown-up.
No matching stories.