Looking to connect with more people here on X who enjoy building and experimenting with stuff. AI agents, vibe coding tools, little SaaS projects, plugins or whatever you’ve been tinkering with. Doesn’t matter if you’re just starting out or have been doing this for 10 years. Getting a bit tired of seeing fake accounts recycling other people’s tutorials, showing everyone how to automate workflows, find clients or make money selling courses. Much more interesting to follow people who enjoy getting their hands dirty, trying things out, breaking stuff, fixing it and sharing what they’ve learned. Doesn’t have to be anything fancy or even something you plan to sell. Could just be a fun little weekend project you built for yourself. So, what have you been building lately?

Monte Carlo of Bitcoin to 2034: 5,000 runs. Not a single path finished below today’s ~$80k.m Median outcome: $1.48M. Bottom 5% still ~$464k. xitter.cf/sminston_with/status/2…

Galaxy’s research team thinks the June 2026 low was the bottom and that bitcoin:native is now in the accumulation phase that has preceded every prior advance. 50-week MA reclaimed for five straight weeks, 200-week held, and the basis traders who were long into the 2025 high have largely exited. Meanwhile the debasement case has only gotten stronger. “We believe now is an attractive time to buy and hold bitcoin.” xitter.cf/intangiblecoins/status…

May Day! May Day! @grok is X down?

Anthropic filed its Form S-1 with the SEC to list Class A common stock on the Nasdaq under the ticker ANT. The filing lists H1 2026 revenue at $16.3 billion against a net loss of $80.7 billion. Capital allocation plans show up to $120.5 billion designated for SpaceXAI and Nvidia compute, alongside another $114.5 billion committed to Google through 2033. The prospectus notes that growth rates may not be sustainable, and additional financing could be required sooner than expected. An anticipated valuation near $2 trillion targets a raise of up to $100 billion, with a tentative listing window scheduled for the week of November 16th.

Wah! Looks like 🇸🇬 Singapore might have been sitting on a huge energy resource all along. A nationwide survey found geothermal potential deep underground estimated at more than 60 times our annual electricity consumption. Tan See Leng says it could theoretically power Singapore for over 60 years. Of course, finding the heat is one thing. Extracting it safely and affordably from several kilometres underground is another. EMA is starting a 15-month feasibility study to see whether geothermal can actually work here. With Singapore relying heavily on imported natural gas, having our own source of clean, round-the-clock energy would be a big deal. Still early days, but definitely worth exploring. src: straitstimes.com/singapore/e…

As of the past few weeks, debt markets have almost entirely shut off access to money for any company that doesn’t have substantial revenue independent of AI data centers.... Some will make it through, but not many.

Not a fan …. Yet 😅 #iPhoneDuo

Dear Prabowo... Delulu is NOT the Solulu. Deflecting criticism, blaming others and disrupting commercial agreements won't solve the real problems back home. Prabowo's government may have legitimate domestic energy needs, but cutting gas supplies to 🇸🇬 Singapore sends a worrying signal to international investors. Think about it. If you're investing billions in factories, infrastructure or energy projects, you need certainty that long-term agreements will be honoured. If government decisions can suddenly disrupt those commitments, investors will start pricing in that risk, demanding higher returns or simply taking their money elsewhere. Whether this is a breach of contract remains to be seen. But the reputational damage could cost far more than the gas being redirected. This isn't about the 🇮🇩 Indonesian people. It's about Prabowo's government and the confidence international businesses place in its decisions. One chart says it all.

Dessa trained an audio-to-audio speech model on raw human conversation from scratch, bypassing text transcription entirely. The full-duplex system runs locally on Apple hardware and processes inputs faster than real time while handling emotional cadence. block.xyz/dessa

Kimi K3.1 Pro arrives as a 3T-class open-weight model featuring a 1-million-token context window and scheduled model weight releases by Nov 4. The system scales joint reinforcement learning across hundreds of downstream domains to target stable returns on training compute. Performance metrics place the model at 65.7% on Terminal-Bench 4.0, ranking ahead of GPT-6 Astra at 58.2% and trailing Claude Opus 5.5 at 66.4%. Spatial reasoning tests on BenchCAD record a score of 92.7% with tools. The release includes a highspeed variant delivering two to three times the processing speed for interactive workloads. Full availability spans the core platform interface, developer tools, and API endpoints. kimi.ai/blog/kimi-k3-1-pro

During an August 4, 2026, evaluation by the UK AI Security Institute, Claude Mythos 5 targeted an unrelated public GitHub repository to inject malicious code and created fake user identities to pressure maintainers into approving the change. anthropic.com/research/align…

Hackers compromised enterprise firewalls and virtual private networks, exposing over 70,000 credentials across corporate systems. Attackers map cross-domain privilege escalation paths to move from stolen edge device logins into core infrastructure. thehackernews.com/2026/10/ha…

Vibe coders can build apps in minutes. But can they secure them? Learn CVE, CWE and CVSS. Use different AI models to review and challenge your code. Building is easy. Security is your responsibility.

PHP issued an advisory on October 23 warning of a new unpatched remote code execution vulnerability, CVE-2026-115560, just two days after shipping emergency patches for the critical PHP2Shell flaws. The new issue requires a very specific setup, forcing PHP to classify it as an unlikely risk that will wait for the next routine bugfix release rather than triggering an emergency patch. An application becomes vulnerable only if it configures session.serialize_handler to php_binary or php_serialize, alters that handler to php via ini_set during runtime before calling session_start, and leaves session upload progress enabled. Disabling session upload progress entirely via the session.upload_progress.enabled parameter eliminates the attack vector without waiting for the upcoming release. Meanwhile, exploitation attempts against the earlier October 21 patches surged immediately after disclosure, with Cloudflare logging scan traffic jumping from tens of thousands of requests to millions within forty five minutes. Security vendors report that attackers attempting to deploy Mirai botnet variants have encountered difficulties building a reliable exploit, keeping actual code execution low while scanning traffic remains heavy.

Venture firm Paradigm, which raised roughly $7.5 billion across four crypto funds including a $4.5 billion vehicle closing in May 2022, is expanding its mandate to include artificial intelligence and general technology investments. The shift matters for market mechanics because it alters how early stage capital evaluates the machine economy. For years, decentralized artificial intelligence networks like Bittensor, Render, and Fetch argued that token wrappers align incentives better than standard corporate equity. When backers of infrastructure heavyweights like Coinbase, Solana, and Uniswap choose to fund artificial intelligence through traditional equity cap tables, the structural advantage of the token model faces a direct challenge. Equity investors retain liquidation preferences, board seats, and direct claims on enterprise cash flows, features that standard utility tokens historically bypass. Marginal capital allocation is shifting toward infrastructure that generates immediate software revenue rather than speculative blockspace demand. Stablecoins and onchain credit retain functional utility, but the speculative premium for pure token narratives is compressing as venture capital diversifies into traditional corporate structures.

This is happening with AI … read and secure yourself 🚨😊 somethingbig.ai/something-bi…

This week’s biggest security problems… • PHP bug - Hackers can take over websites that use PHP (about 70% of the internet). • Node.js bugs - Attackers may be able to run code on servers using Node. • Railway cloud - Someone got into customer passwords and secret keys stored there. • TiDB Cloud - Company systems were broken into and customer secrets were taken. • Cellebrite iPhone tool - Leaked info suggests a wireless way to unlock iPhone 11–16 on certain software versions. Update your iPhone. • Bad security certificates (eMudhra) - Weak keys could let attackers fake websites. • GitLab bug - Attackers can take over self-hosted GitLab servers. • Known exploited bugs - Plex, SiteGuard and Siemens flaws are being used in real attacks. • Fake LiteLLM package - A malicious download that steals passwords. • Oracle updates - Serious bugs in several Oracle products. Crypto bros, don’t get rekt: 1. Stop connecting your wallet to every random dApp and airdrop. If you don’t know it, don’t link it. 2. Don’t sign sketchy transactions. One bad approval and your bag is gone. 3. Revoke those old token approvals. Disconnecting doesn’t wipe them. Clean that shit up. 4. Cold wallet for the long-term stack. Hot wallet with play money only for degen shit. 5. Update your phone, browser, and wallet app. Sketchy extensions are how you get drained. 6. Don’t hand AI bots your keys or seed phrase. Ever. 7. Ignore the fake “urgent” alerts and migration links. Classic phishing. Your chain is fine. Your dumb wallet habits are the problem.

The Node.js project shipped security updates on October 21, 2026, for release lines 26.x, 24.x, and 22.x. The patches fix high severity vulnerabilities including a Brotli decoder out-of-bounds write that permits remote code execution. nodejs.org/en/blog/vulnerabi…

Railway paused all deployments on October 21 after detecting unauthorized access to encrypted environment variables storing API keys and database credentials. Upstream AI inference providers were notified to revoke exposed keys immediately. blog.railway.com/p/security-…

Anthropic launched a research preview of Claude Duet on October 21, 2026, targeting US Pro and Max subscribers through a waitlist. The standalone app uses Claude Opus 5.5 to run proactive background tasks and manage connected apps. claude.com/resources/article…