Avoid Tax Season Scramble
Don’t wait until spring to scramble through deductions, documents, and expenses. BELAY’s experienced tax prep professionals can help you get organized before it turns into an emergency.
Download the free Personal Tax Prep Checklist to start today.
Today In Ai
3 things that happened while you were busy
1 One of the largest AI-focused hedge funds sold its entire public portfolio to Citadel after a leveraged bet on AI infrastructure unwound fast.
Leopold Aschenbrenner, the 25-year-old former OpenAI employee turned hedge fund manager, has reportedly sold Situational Awareness's entire public portfolio to Ken Griffin's Citadel. His strategy centered on concentrated, leveraged positions in AI infrastructure stocks — a bet that produced extraordinary returns while the market climbed, then reversed hard during the past month's rapid selloff. The fund's rise and fall in under two years is a compact case study in what leverage does in both directions.
2 Freehand raised $75M for an AI platform that hunts down overpaid corporate invoices most finance teams don't have time to chase.
Freehand's system monitors the entire invoicing process — double-checking contracts, messaging suppliers directly, and reconciling data against agreed terms. The startup claims it recovered $260M in overpaid invoices last year across clients including Meta, Johnson & Johnson, and Unilever, by tracking down small discrepancies that most companies simply do not have the bandwidth to pursue individually. See how it works.
3 Stan launched Stanley, a texting agent that writes and posts your social media content while stripping out the tells that make writing sound AI-generated.
Stanley plugs into your most-used tools, learns your specific writing voice, and proactively drafts posts for LinkedIn, X, Instagram, and other platforms. The startup says every post is run by you before it publishes, and Stanley is built to automatically eliminate common AI writing tells — em dashes, "it's not x, it's y" constructions, and similar patterns readers have learned to spot. Watch Stanley in action.
From The Frontier
Four stories from this week that will still matter in six months.
The pacing warning. Over 1,200 employees at frontier AI companies signed an open letter cautioning that AI could advance "beyond our ability to understand or control the resulting systems," and calling on the US government to coordinate deliberate pacing at the international level to ease the competitive pressure each individual lab feels. Signatories include chief scientists at OpenAI, Anthropic, and Meta. The signatory list spanning companies and seniority levels is itself the signal — this is not outside criticism, it is insiders speaking on the record.
The price shift. Anthropic's Claude Opus 5 matches Fable 5 on many benchmarks at roughly half the price, live now across all Anthropic platforms, with a fast mode running at 2.5x normal speed. Anthropic also published an Opus 5 Prompting Guide to accelerate productive use from day one. Enterprise teams that were rationing frontier-model usage because of cost now have a materially cheaper way to get comparable results.
The open-weight jolt. Moonshot AI open-sourced the full weights for Kimi K3, now the largest open-weight model ever released, cleared for research, personal, and commercial use with some deployment guardrails. Demand was intense enough that the lab paused new subscriptions to manage compute load in the weeks before the weights went public. K3's quality and price will put direct pressure on how the top US labs justify their own pricing.
The voice unlock. Fish Audio publicly launched S2.1 Pro, its flagship voice model, with founders demoing the technology by cloning their own voices live in real time. The company closed $52M in seed funding to cap off its first year. Between Stanley eliminating AI writing tells and Fish Audio making voice cloning instant, the tools for producing content indistinguishable from human output are consolidating fast.
In The Know
What people are actually watching and sharing
✏ Improved writing. A simple prompt that eliminates most signs of AI writing has racked up 11,000 bookmarks. Upload it before every conversation, or add it permanently to your instruction file, and the improvement compounds across everything you produce afterward.
😅 The big leagues. A five-step interview process can feel excessive — until you compare it to OpenAI's. One candidate leaked what the process actually involves (1M views). Given OpenAI's median compensation, most commenters agreed the time investment is worth it.
🧐 Seems sloppy. LinkedIn is testing a "seems like AI slop" reporting button that lets users flag posts that simply feel off, whether or not they can prove it. Given the platform's current reputation, expect the button to see heavy use. See which platforms suffer most from AI slop overall.
👀 Uncanny valley. A Redditor asked ChatGPT to generate an image from "the deepest corner of its mind" and got something they said they loved and hated in equal measure (1,900 upvotes). Worth seeing for yourself rather than taking the description on faith.
🤳 Minimizing screens. A dad wanted his toddlers to talk with their grandparents whenever they wanted, without handing them a phone. He vibe-coded an out-of-the-box solution that a large number of other parents are now asking to replicate (1M views).