Kimi K3 Released as Largest Open-Source Model with Low Caching Cost

Kimi K3 released as largest open-source model (2.8T params, 104B active) with Delta Attention for low memory and cost ($0.30 per million tokens) — via 1
Digital M&A market sees massive activity: Flippa reports 90% of exits executed poorly; acquire.com lists multiple businesses for sale (TTM $7.4k–$390.8k); Codie Sanchez’s team screened $90M in LOIs — via 1 2 3
Indie developers face AI headwinds: @levelsio claims SaaS is dying and traffic/revenue declining; but @starter_story showcases counterexamples (Sarah $340K/month in 60 days, Jack $147K total) — via 1 2
AI agents automate enterprise tasks: Codex sales assistant can query codebases, fill FAQs, create S3 configs autonomously — via 1
2026 bonus depreciation returns to 100%, enabling major tax deferral through cost segregation — via 1

Kimi K3 open-source model: Largest open-source model (2.8T parameters, 104B active) introduced by Moonshot AI. Uses Kimi Delta Attention to reduce memory consumption and speed up long-context processing. Caching cost is extremely low at $0.30 per million tokens, emphasizing context compression and precise retrieval. — via 1
AI agent automation: Codex enterprise sales assistant demonstrates ability to search codebases for compliance data, auto-fill FAQs with uncertainty flags, and create S3 configurations. Arvid Kahl argues that agent systems require reliable delegation; human “quick fixing” is unsustainable because agents have more memory, patience, and rationality to follow instructions. — via 1 2

M&A market data and opportunities: Flippa reports that 90% of business exits are executed poorly, eroding founder cash via taxes, bad planning, and restrictive non-competes. It highlights a resale marketplace for sale: $34.9K monthly profit, 70–100% annual growth, $1.5M ARR, 3.6M monthly visitors, tied to circular economy and ESG trends. acquire.com lists multiple startups for sale (Shopify app, AI health app, AI SaaS, eCommerce) with TTM revenue from $7.4k to $390.8k and asking prices from $35k to $1.1M. Codie Sanchez seeks 5 buyers for businesses over $2M; her team has screened over $90M in LOIs in 4 months. — via 1 2 3 4
M&A process advice: acquire.com emphasizes that phone calls during due diligence resolve questions in real time and prevent deal decay from email delays. For sellers, a clear process and real buyers are critical to converting conversations into deals. — via 1 2

SaaS model under threat: @levelsio asserts that “SaaS is dead” and most subscriptions can be canceled with Vibecode It. He observes a trend of declining traffic and revenue for indie developers, attributing it to large AI models eating the application layer. (Disputed/unverified: no supporting data.) — via 1 2
Counterexamples of success: @starter_story shares two case studies: Sarah built an app to $340K/month in 60 days by focusing on content and distribution before product; Jack launched Profit AI and reached $147K total revenue in 5 months, with app MRR at $30K and service MRR near $40K. — via 1 2
China’s speed and market perception: @levelsio warns that the West underestimates China’s speed and capability, citing a before-and-after comparison from 18 months ago, and recommends visiting Chinese cities to understand the hospitality and progress. — via 1 2

Discover