diff --git a/_posts/2025-12-31-codatta-2025-reflections-and-2026-forge.markdown b/_posts/2025-12-31-codatta-2025-reflections-and-2026-forge.markdown
new file mode 100644
index 0000000..4292c49
--- /dev/null
+++ b/_posts/2025-12-31-codatta-2025-reflections-and-2026-forge.markdown
@@ -0,0 +1,42 @@
+---
+layout: post
+title: 'Codatta 2025: Reflections and 2026 "Forge"'
+date:   2025-12-31
+categories: [Ecosystem & Community]
+author: yi_zhang
+status: official
+tags: [community]
+excerpt: "Looking back at what we built together in 2025 — and how we plan to forge the next chapter of Codatta in 2026."
+---
+
+## Codatta 2025: Reflections and 2026 “Forge”
+
+## Introduction
+In 2025, we laid the groundwork for Codatta after [two years of building](https://docs.codatta.io/en/community/history): we launched $XNY (TGE, Binance Perps Listing) to the market and openly pressure-tested our protocol, all while refining our broader vision. Our primary focus is addressing the accelerating need for **expertise data** in AI—high-signal, verifiable knowledge that models can reliably learn from and enterprises can trust. This sets the stage for 2026: turning distribution into delivery.
+
+## Look-back: what we achieved and what’s still unproven
+In 2025, $XNY (Codatta’s native token) entered broad circulation via major exchanges (e.g., [Binance](https://www.binance.com/en/support/announcement/detail/901d960af2134457aa1e9d72ce6e30fd), [Kraken](https://x.com/krakenfx/status/1948488362983067985), [Bybit](https://www.bybit.com/en/price/codatta/), etc.), with special thanks to the [Binance Booster program](https://www.binance.com/en/support/announcement/detail/901d960af2134457aa1e9d72ce6e30fd) (Codatta as the first project), which gave us distribution and attention that most infrastructure projects never get this early.
+
+We also crossed **[1M+ KYC'ed users](https://x.com/codatta_io/status/1976288857177149590)**, which is a real starting line for enterprise-grade data; importantly, we used distribution moments (e.g., [Binance Booster](https://www.binance.com/en/support/announcement/detail/901d960af2134457aa1e9d72ce6e30fd)) to recruit and train contributors rather than chase empty metrics.
+
+Commercialization started with stable PoV workloads (e.g., [**CipherOwl**](https://cipherowl.com/)), but the proof we still owe is harder: [**TNPL (Train-Now, Pay-Later)**](https://docs.codatta.io/en/core-systems/royalty-engine#tnpl-train%E2%80%91now%E2%80%91pay%E2%80%91later) must work end-to-end (metering → lineage → settlement → royalties) without manual negotiation, and we must build a frontier expert talent pool that produces datasets that measurably move model performance (not just more labels)—which is exactly what our [protocol overview](https://docs.codatta.io/en/quick-guide/overview) is designed to enable.
+
+## Macro: what changed around us
+On the AI side, people started to challenge hype and bubble in AI industry, again. The reality is that AI capability compounds: transformer-based foundation models (LLMs, and multimodal variants like VLM/VLA) may still be far from AGI, but they’ve already delivered irreversible, high-leverage impact—while GPT/Gemini/Qwen keep improving, the frontier has moved to “how we build” (LLM-native product patterns, agent workflows, and protocols like Claude Skills and MCP servers), and tools like Claude Code/Cursor/Gemini3/ are the everyday proof that coding is not going back to what it was (see [Fei-Fei Li on why data matters](https://computerhistory.org/blog/fei-fei-li-talks-ai/), Karpathy on [“vibe coding”](https://x.com/karpathy/status/1886192184808149383), and the emerging “ship by system” reality in posts like [Claude Code in the wild](https://x.com/karpathy/status/2005067301511630926)).
+
+On the crypto side, mainstream acceptance is increasingly institutional and policy-shaped—[spot Bitcoin ETP approvals](https://www.sec.gov/newsroom/speeches-statements/gensler-statement-spot-bitcoin-011023) put “Wall Street rails” under BTC exposure, stablecoins are being pulled into mainstream settlement (e.g., [Visa stablecoin settlement](https://investor.visa.com/news/news-details/2025/Visa-Launches-Stablecoin-Settlement-in-the-United-States-Marking-a-Breakthrough-for-Stablecoin-Integration/default.aspx)), Ethereum keeps scaling toward real-world throughput (e.g., [Dencun](https://www.investopedia.com/what-you-need-to-know-ahead-of-ethereum-dencun-update-wednesday-8607518)), tokenized funds like BlackRock’s **[BUIDL](https://www.businesswire.com/news/home/20240320771318/en/BlackRock-Launches-Its-First-Tokenized-Fund-BUIDL-on-the-Ethereum-Network)** are live, and [prediction markets](https://www.ft.com/content/bc300579-4b19-4f82-b232-e537e992c310) are entering mainstream conversation; the disappointment is that outside these finance-adjacent lanes, “innovation” still clusters around DeFi and gambling-like speculation, while narrative-first projects dominate without sustained delivery (often summarized as [“altcoin season is on hold”](https://www.coindesk.com/markets/2025/11/18/bitcoin-is-bleeding-but-market-observers-say-altcoin-season-is-on-hold)). So the bar must be higher: don’t ship tokens, ship systems that deliver repeatable value.
+
+## Look-forward: 2026 “Forge”
+Our 2026 theme is **[Forge](https://docs.codatta.io/en/community/history)**: make the hybrid protocol work end-to-end and plug into real businesses.
+
+Why now: [ImageNet proved that data can unlock algorithmic leaps](https://computerhistory.org/blog/fei-fei-li-talks-ai/), and demand is shifting from labor-intensive basic labeling toward expertise data (reasoning traces, domain judgments, adjudication, high-signal evals), which makes ownership and attribution not ideology but inevitability.
+
+What we will ship in 2026 is deliberately practical, executed as three loops: **(1) protocol loop (v1)**—automated screening + anchoring, hybrid delivery (decentralized source-of-truth + cloud hot paths + secure compute), access control & metering v1, ownership & royalties v0.x, and real $XNY utility in production paths (see the [protocol overview](https://docs.codatta.io/en/quick-guide/overview)); **(2) network loop**—turn 1M+ KYC into an activated quality engine via an identity ladder, expert tracks, 20+ regional pods, and evidence-heavy qualification/QC that enterprises can audit; and **(3) business loop**—scale commercialization (not pilots) by expanding from today’s proven workloads into repeatable verticals (healthcare, robotics) and closing lighthouse customers that force real-world reliability.
+
+Stage story: a startup hits a Codatta endpoint, receives a 402, pays, retries, and the signed usage event is metered, tied to a dataset version and policy; later, the Royalty Engine replays the same events and pays contributors/validators/backers automatically—lineage resolves disputes by construction (again, see the [protocol overview](https://docs.codatta.io/en/quick-guide/overview)).
+
+## Closing
+
+2025 proved distribution + verified starter; 2026 must prove delivery: turn expertise into data assets, connect them to production usage, and route value back to creators reliably and repeatedly at scale.
+
+— Yi Zhang (PhD, CEO & Founder)
\ No newline at end of file