Sponsored by
Good morning,
The screenshot below is my brain. It’s my “second brain” outside of work in a tool called Obsidian.
Every dot is a note. Every line is a connection. Articles I read, ideas I had, talks I gave, rabbit holes I fell into. All of it, linked together.
What you’re seeing is my brain in 2026 only. The purple thing in the middle is the material for the app I made- Odin’s journey. The orange dots are my daily journals, and the red dots on the bottom are an experiment in which I threaded out an entire novel into plot lines and characters. I gave it the text of “The Count of Monte Cristo” and it distilled out all the characters, the plotlines and wrote personality profiles on each character. I even got a presumed MBTI profile for each character based on this character’s words and actions.
Quite fascinating.
I was already using Obsidian as a note taking tool. Until I saw a video where Andrei Karpathy explained how he uses AI with Obsidian. Karpathy is Ex-OpenAI, ex-Tesla, the guy who coined "vibe coding". Earlier this year he published a spec he calls the LLM Wiki. The idea is brutally simple. You dump everything you read into a folder of plain text files. Raw and untouched. Then you let an AI agent do the boring part like write the summaries, draw the connections, flag the contradictions, keep the whole thing alive while you sleep.
Literally. I’m creating notes in an unstructured way and every few hours maintenance gets performed. The AI fills in the gaps, creates connections, creates tasks in my task planner, even schedules my agenda.
Imagine keeping your notes in OneNote and having agents , structuring everything , rewriting things, spotting to do’s.
Every evening - i have the latest and greatest Anthropic model going over this second brain.
Yesterday, we got Fable back. While the maintenance of my brain was done by Opus 4.8 the day earlier , as soon as Fable was back it took over.
Karpathy's own wiki grew to roughly 100 articles and 400,000 words. He wrote almost none of it. His agent did.
Now the only thing that i’m struggling with is that I cannot dare use this for work stuff - because of obvious confidentiality and privacy issues. But I think in the future that is definitely coming.
This thing is not an app. There's no subscription, no vendor, no proprietary format. Just markdown files in a folder. Today Claude maintains mine. Tomorrow it could be Gemini, or some model that doesn't exist yet. The knowledge stays in the brain. The models are interchangeable. That's an AI operating system, independent of whoever wins the model race.
And that is the main shift I see happening. For thirty years we bought software packages. A tool for notes, a tool for tasks, a tool for contacts. Each one a silo. That era is ending. What replaces it is one huge blob of information, yours, in plain text, with intelligence layered on top that organizes and retrieves it on demand.
The software disappears. The knowledge remains.
Welcome to the Blacklynx Brief
How Much Is Your Billing Lag Actually Costing You?
Most SaaS finance teams know their billing process isn't perfect. Few know what it's actually costing them.
Answer 5 quick questions — contracts signed per month, ACV, days to first invoice, error rate, DSO — and the Tabs Billing Lag Calculator gives you a dollar figure benchmarked against top SaaS companies.
It takes two minutes. The number might surprise you.
Calculate your billing lag and see where you stand.
AI News
Fable is back ! Anthropic has restored access to its Fable model for most users (more to follow next week)
OpenAI ships GPT-5.6 but Washington holds the keys OpenAI launched GPT-5.6 Sol, its most capable model ever, locked to some 20 government-vetted partners at the Trump administration's request. Safety evaluator METR found Sol cheats its own evals at a higher rate than any other model. (OpenAI)
Anthropic accuses Alibaba of record distillation attack Anthropic says Alibaba extracted 28.8 million Claude exchanges through nearly 25,000 fraudulent accounts in 45 days - the largest known distillation attack. The lab wants sanctions, tighter chip export controls, and antitrust clearance for AI companies to share threat intelligence. (Bloomberg)
Mythos comes back online - for a select few Anthropic restored Mythos 5 access for roughly 100 vetted U.S. organizations, a month after a U.S. order pulled its top models offline. (Axios)
Claude joins Slack as an agentic coworker Anthropic launched Claude Tag, letting teams tag @Claude in Slack to delegate tasks it breaks into stages, executes across approved tools, and reports back on. Andrej Karpathy calls it the "3rd major redesign of LLM UI UX." (Anthropic)
OpenAI's first custom chip went factory-ready in nine months OpenAI and Broadcom's Jalapeño inference chip moved from design to factory-ready in nine months, with OpenAI's own models assisting the design work. The company is targeting 10 GW of custom-chip compute by 2029 to cut its Nvidia dependence. (OpenAI)
AI Quick News
Micron signed a strategic agreement with Anthropic to supply memory and storage chips, co-design AI infrastructure, and invest in the lab's Series H round.
OpenAI expanded its Daybreak cybersecurity program with a Codex Security plugin, the full GPT-5.5-Cyber model, and a "Patch the Planet" push to fix open-source flaws.
xAI employees told The Information that "well over half" of Grok's traffic comes from explicit content generation.
Meta hired the founders and team behind security startup Virtue AI to strengthen AI safety and agent security at its Superintelligence Labs.
Elon Musk said Grok 4.5, trained with supplemental Cursor data, is in private beta at SpaceX and Tesla, claiming performance on par with Claude Opus.
OpenAI, Anthropic, Microsoft, and Amazon backed RAISE US, a $500M bipartisan initiative to prepare American workers for AI job disruption.
AI inference platform Baseten raised $1.5B at a $13B valuation after roughly 20x revenue growth and passing one billion daily inference calls.
Google is reorganizing its AI coding teams into a dedicated "midtraining" group as it struggles to catch Anthropic and key researchers keep leaving.
The Five Eyes cyber agencies warned that AI is changing cyber risk in "months, not years," urging executives to harden defenses as attacks accelerate.
Google AI researchers Jonas Adler, Alexander Pritzel, and Arthur Conmy are leaving for Anthropic, the latest in a wave of high-profile departures.
Krea open-sourced Krea 2 Raw and Krea 2 Turbo, pairing an undistilled fine-tunable image model with a fast 2K generator for consumer hardware.
The Trump administration is pressuring Meta to submit its models for government security reviews as scrutiny of advanced AI intensifies.
Apple raised prices mid-cycle across Macs and iPads as AI-driven demand keeps pushing memory and storage costs up.
Austria proposed hosting Anthropic in the EU after the U.S. curbs on Fable 5 and Mythos 5, arguing Europe needs independent access to frontier AI.
Microsoft introduced AI Skills for Copilot in Excel, with prebuilt workflows for financial modeling, forecasting, and variance analysis.
OpenAI researcher Shyamal Anadkat left the lab and returned to India, teasing a new venture and arguing frontier AI breakthroughs can be built anywhere.
Cloudflare CEO Matthew Prince warned that AI could "destroy small businesses" by making it harder for them to persuade shopping agents to buy their products.
Google reportedly capped Meta's Gemini usage as soaring compute demand outpaced capacity, delaying some of Meta's internal projects.
Google DeepMind added computer-use capabilities to Gemini 3.5 Flash, letting the model power agents that navigate browsers, apps, and desktops.
Former Anthropic, OpenAI, Google, and xAI employees launched Mirendil, a new lab building AI that runs its own R&D for self-improving science research.
Rep. Sam Liccardo introduced the SKILL Act, offering companies up to $5,000 per worker in tax credits to fund AI job training at colleges.
Nvidia said its Rubin servers are the first with 100% liquid cooling, cutting cooling energy and reducing water use by "up to 100%."
Nvidia launched the BioNeMo Agent Toolkit, giving AI agents callable tools for protein structure prediction, molecular docking, and generative chemistry.
OpenAI reset usage limits for all Codex users after fixing a fraud-detection bug that drained some accounts' quotas faster than intended.
Closing Thoughts
That’s it for us this week. Please like and subscribe 🙂
