Teaching Claude to tell a completed deal from a bid: buyers matched 192/193, deal status stays hard
YvesMulkers · x · 2026-11-21
7wData shares a practical method for making Claude judge the true status of M&A headlines: two models independently reading the same "Fox plans to acquire Roku for $22B" story gave different answers, and the disagreement traced to a single verb, "plans."
The distilled framework asks two questions of every deal story: who is the buyer, and how far has the deal gone. The first is easy; the second lives in the verb.
Tested on 2026 takeover stories collected for their research platform Ins7ghts:
- For completed or agreed deals, two independent readings named the same buyer 192 times out of 193
- Distinguishing agreed vs completed, or bid vs agreed, did not hold up reliably — the authors flag this openly
They map status to typical phrasing: completed ("Salesforce completed its acquisition of Informatica"), agreed ("Nielsen agreed to acquire DoubleVerify at $13.60/share"), bid ("Schneider Electric plans a takeover offer for Shelly Group at €70/share"), and stake (Meta's $14.3B for 49% of Scale AI). Prompts included.
More from coding & agent
- AI gateways are becoming the key control layer in production AI stacks: 8 capabilities — goyalshaliniuk · 2026-11-21
- Mnemos 2.0: agent memory designed to preserve what matters for the next session — RileyRalmuto · 2026-11-21
- InferenceBench ranks agents on LLM serving speedup: Claude Opus 5.5 tops at 12.08x, agents lose to simple hyperparameter search — maksym_andr · 2026-11-21
- 4 architecture patterns to tame multi-agent system reliability, from a 27-year reliability engineer — bibryam · 2026-11-21
- AutoCut + FrameAtlas: open-source local-first video editing driven by Claude Code via MCP — moheshmohan · 2026-11-21
- InfoQ: What an agent harness is, and two ways to build one — bibryam · 2026-11-21