Seb

Everyone tells you that agent memory vaults are important BUT nobody shows you HOW to do it. Copy paste below into your Claude code, ask it to investigate your current system, think, and build a plan for your own Second Brain memory vault in Obsidian. ______________ SecondBrain + SB Vault Architecture Two local repos working together: 📂 SecondBrain - my executive operating system - Context files (priorities, tasks, team, goals) - Active project folders with strategy docs - AI agents (Tom = Telegram bot, Jerry = research, Cindy = content) - Decision log (append-only) - Skills system (slash commands for repeatable workflows) - Claude Code as the primary interface 📂 SB Vault - Obsidian-compatible knowledge wiki - Three layers: raw/ (immutable sources), wiki/ (AI-generated pages), schema (CLAUDE.md) - Wiki has: sources/, entities/ (people, orgs), concepts/, synthesis/ - Every page has YAML frontmatter (title, type, dates, sources, tags) - Uses [wikilinks] for cross-references - hot.md = rolling context (max 50 lines, auto-pruned after 14 days) - index.md = master catalog of all pages 🔄 *How they connect* - One-way sync: SecondBrain → SB Vault (not bidirectional) - Runs daily at 07:03 via Tom's scheduler (node-cron) - Three data sources feed the sync: 1. Queue file (items appended during work sessions) 2. Decision log (new strategic decisions auto-detected) 3. Jerry briefs (research outputs dropped in raw/) - Sync creates/updates entity pages, logs to hot.md, updates index - Guardrails: max 5 entities + 1 brief per run, 15min timeout, never creates synthesis (manual only) 🧠 The logic - SecondBrain = operational layer (what's happening NOW) - SB Vault = knowledge layer (what we KNOW, compounding over time) - Agents work in SecondBrain, their outputs flow into the vault - The vault is queryable context for future decisions Stack: Node.js, local filesystem, Obsidian-compatible markdown, Claude Code as the AI layer. No cloud DB.

new life-changing feature just dropped on @claudeai in VSC and it immediately destroyed a couple of ai start-ups with just this 1 app update TLDR you're dead if you're building things that's AI-adjacent. these companies have infinite resources and can crush you anytime the answer is to build things/tools/infra that AI companies can't build in and niche it down, and that is still a huge BLUE ocean

.@decentra1ized_ from @SolanaFndn on growing as a Solana builder in APAC. Talks from @jemmmyjemm (@MonkeDAO), @jacobkxyz (@Superscrypt), @Sirolaso (@MesaStudio_so) and @TalentCorpMsia.

@SuperteamMY @solana 🍻

The arena was packed 🌐👾 The crowd grew. The pitches landed. The builders showed up. @solana Day @ Solana Network State Spring 2026

@solana @tuakdotsol @superteam @ns @decentra1ized_ @jacobkxyz @jemmmyjemm @TalentCorpMsia @SuperteamMY 👁️👄👁️

@tuakdotsol @superteam @ns @decentra1ized_ @jacobkxyz @jemmmyjemm @Sirolaso @TalentCorpMsia @SuperteamMY 👀

are y'all ready for @solana day with @superteam! wear your shirt and come by to the @ns coworking space to pitch and hear from our partners @decentra1ized_, @jacobkxyz, @jemmmyjemm, @Sirolaso and @TalentCorpMsia. today's merch drop are these dope @SuperteamMY pins. the starter pack is complete - which is your fave? ;)

Think CHIP is pretty clearly overvalued here, tho it depends if you’re looking at FDV or mcap. Some napkin math. Assuming @USDai_Official successfully executes its $665M loan pipeline ($61M today), that’s ~$24M in annualized revenues: (+) $98M (gross interest at 15% APR) (+) ~$5M origination fees + T-bills on idle buffer (-) $73M to sUSDai holders (11% APR) (-) $3.3M insurance premiums to Barker (0.5% on loan book) (-) $5M operating costs At a generous 20x early stage multiple = $500M valuation. Seems like a steal at $200M mcap – you’re buying at 40% of fair value for a future full loan book activation with a reasonable margin of safety. But FDV is a lot more relevant than mcap here. CHIP’s cliff ends in Q1 2027 when investor (29.6%) and core contributor (23.5%) tranches begin unlocking. Crypto tends to frontrun unlock events ~6 months ahead. So CHIP’s repricing window effectively starts Oct 2026, ~6 months from now. At $1B FDV today, you're paying a price that requires active loans growing from $61M today to ~$1.3B. That’s a tough 21x execution hurdle within a short time, even in the most bullish AI capex market. Not a knock on USDAI btw. Genuinely think it’s a great team with proven stellar execution (Paypal, Munich Re, multiple CEX listings). Just overvalued right now and not a long-term bet i’d take until after post-unlocks repricing.

Our @superteam perk recipients and @colosseum Frontier participants have been building in stealth. Tomorrow is their moment, first showcase of the season at @solana day (📍@ns). Watch them pitch to a bigger crowd and real ecosystem partners. See you there 🫡

@notthreadguy @cobie @grok think and tell me whats the top 5 take aways from this interview

my full conversation today with @cobie 01:04 offering to work at Coinbase for free 06:05 the K-shape crypto thesis 10:33 thoughts on Saylor 14:39 why UpOnly hasn't come back 22:37 the last 10 years of Crypto 44:52 trillion dollar IPOs 57:02 Cobie's legendary buy wall 1:02:57 top 5 Crypto traders of all time 1:21:20 reasons to be optimistic on Crypto

ngl loving this new @superteam merch coming your way for Solana Day this Friday at @ns! @decentra1ized_ from @SolanaFndn, @jacobkxyz from @Superscrypt, @jemmmyjemm from @MonkeDAO, @Sirolaso from @MesaStudio_so and @cradlefund will be here to watch you pitch! get the merch drop, we have buses to help you come to NS too. deets in 🧵

@arbitrum basically the chain runs on excel

The Arbitrum Security Council has taken emergency action to freeze the 30,766 ETH being held in the address on Arbitrum One that is connected to the KelpDAO exploit. The Security Council acted with input from law enforcement as to the exploiter’s identity, and, at all times, weighed its commitment to the security and integrity of the Arbitrum community without impacting any Arbitrum users or applications. After significant technical diligence and deliberation, the Security Council identified and executed a technical approach to move funds to safety without affecting any other chain state or Arbitrum users. As of April 20 11:26pm ET the funds have been successfully transferred to an intermediary frozen wallet. They are no longer accessible to the address that originally held the funds, and can only be moved by further action by Arbitrum governance, which will be coordinated with relevant parties.

Introducing aWETH Redemption Protocol With ETH utilization at 100% on Aave, many lenders are currently unable to withdraw and face increasing risk if markets move. aWETH Redemption Protocol allows ETH lenders to: • Exit into wstETH or weETH • Regain immediate liquidity • Reduce exposure to liquidation risk If you’re just lending ETH — you can fully exit. If you have ETH collateral and another debt — your collateral is seamlessly swapped into wstETH or weETH while your debt remains the same. We’re working alongside @LidoFinance , @ether_fi, @0xProject, @1inch, @KyberNetwork, and other ecosystem partners to: • Reduce systemic risk in DeFi • Ease utilization pressure • Support a healthier DeFi market Our goal is simple: protect users while reinforcing the foundations of DeFi. Capacity is initially limited to $1B in ETH. fluid.io/lite/aave-v3/e…

another day, another exploit

@shu_bit_ great designers will 10x their productivity. good designers will go into ai management. mediocre ones will be phased out.

i am a huge advocate for good designers, and it’s hard to ignore how current ai trends are de-incentivizing them from pushing boundaries. to be fair, there were way too many mediocre designers who frustrated everyone. ai makes those people happy because they can now just bypass the friction and get the work done themselves. but for someone who truly values great design, this trend is depressing for the designers who actually have the potential to lead the industry. it’s destroying the very environment where the next generation of elite talent should be growing. first, designers were already undervalued relative to their impact on user experience. now, as ai drives the cost of replication to near zero, original effort doesn't get properly valued. it just becomes data for the next model to copy. it’s becoming increasingly difficult to swim against this tide. when you can ship fast and rely on quantitative testing for "sticky" results, why fight for innovation? this path leads to products that are addictive, but rarely elegant or meaningful. still, i hope we don't lose those "sparkling" designers who refuse to take the easy way out. even when the cycle gets shorter and more extractive, there’s a human soul in design that ai can’t make originally. i really hope those who stay firm and keep creating something truly new get the recognition they deserve. on a side note, is figma really going to die immediately? or has the stock market just become another meme market like oil is. figma is a tool for communication, not just drawing. just like how some people still choose to trade with their own brains and hands instead of using bots, design taps into that intrinsic human urge for ownership and control. anw this is just my morning ramble and if you read this far, please know that i-si-tteru you

We fixed a bug where rate limits on Claude subscriptions weren't properly adjusted for long context requests in Opus 4.7. We've reset 5-hour and weekly rate limits. Enjoy Opus 4.7!

bruh @claudeai opus 4.7 removed the toggle to expand and view claude's thinking processes massive downgrade man literally unusable I WANT TO SEE IT'S THINKING how else can i know if it's thinking like an idiot or it's doing excess inner monologue debates...

hmm fixing things that were previously hugely complained about kinda sussy but overall a good update

@shawmakesmagic you forgot to audit loop this once everything was done

The quality of your vibecoded slop is horrible. I've seen it. Absolute dogshit. Fortunately, there is a fix. Use this prompt: I want to clean up my codebase and improve code quality. This is a complex task, so we'll need 8 subagents. Make a sub agent for each of the following: 1. Deduplicate and consolidate all code, and implement DRY where it reduces complexity 2. Find all type definitions and consolidate any that should be shared 3. Use tools like knip to find all unused code and remove, ensuring that it's actually not referenced anywhere 4. Untangle any circular dependencies, using tools like madge 5. Remove any weak types, for example 'unknown' and 'any' (and the equivalent in other languages), research what the types should be, research in the codebase and related packages to make sure that the replacements are strong types and there are no type issues 6. Remove all try catch and equivalent defensive programming if it doesn't serve a specific role of handling unknown or unsanitized input or otherwise has a reason to be there, with clear error handling and no error hiding or fallback patterns 7. Find any deprecated, legacy or fallback code, remove, and make sure all code paths are clean, concise and as singular as possible 8. Find any AI slop, stubs, larp, unnecessary comments and remove. Any comments that describe in-motion work, replacements of previous work with new work, or otherwise are not helpful should be either removed or replaced with helpful comments for a new user trying to understand the codebase-- but if you do edit, be concise I want each to do detailed research on their task, write a critical assessment of the current code and recommendations, and then implement all high confidence recommendations.