buildOne report1 publisher Lianxi proposes a gateway that limits a manipulated company agent to what policy allows
Lianxi, an independent project, proposes that a company agent behind Meta and Sierra's PAP hold no credentials and act only through single-use grants. Its authors say the worst case becomes whatever policy allows. The design is a prototype with no independent review.
Reality
- Evidence22
- Adoption
- Not scored
- Hype gap−10
- Incentives35
- Confidence55
buildOne report1 publisher Personal Agent Protocol agents read poppy.json, then verify it against the issuer's OAuth metadata
Sierra and Meta's Personal Agent Protocol, draft 0.1 with 35 design partners, makes /.well-known/poppy.json the first file a user's AI agent reads. The issuer's OAuth metadata must list the domain too.
Reality
- Evidence45
- Adoption3
- Hype gap+12
- Incentives35
- Confidence50
buildConfirmed8 publishers Google's two new speech-to-speech models run tool calls in the background without breaking the dialogue, and Artificial Analysis prices them at $0.84 and $3.50 per hour of input audio on one benchmark subset.
Perspective Coverage
8 publishers
- Builder
- Builder 48%
- Operator
- Operator 29%
- Investor
- Investor 23%
Reality
- Evidence60
- Adoption22
- Hype gap+30
- Incentives72
- Confidence64
Instinct's text-message agent keeps pace with Muse and Dots on household chores
Instinct, the invite-only AI agent valued at $10 billion in late September, held its own against Muse and Dots on everyday chores in The Verge's testing. With results that close, its difference comes down to a text-only interface and a narrower focus.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+20
- Incentives55
- Confidence40
Meta, Walmart, Stripe and Sierra are publishing an open protocol to identify AI shopping agents as Stripe Link's agentic purchases rose 38-fold in a month. Stripe and Walmart also back rival protocols, so merchants setting rules for bots face several standards at once.
Perspective Coverage
12 publishers
- Builder
- Builder 33%
- Operator
- Operator 42%
- Investor
- Investor 25%
Reality
- Evidence60
- Adoption10
- Hype gap+35
- Incentives70
- Confidence62
buildConfirmed2 publishers Meta, Walmart and AI startup Sierra are working on universal standards for AI agents used across business systems. The only account of how it would work comes from a Futurum analyst, who says it holds only if OpenAI and Anthropic adopt it.
Reality
- Evidence50
- Adoption8
- Hype gap+40
- Incentives55
- Confidence55
Repeat buyers push AI-to-AI startup acquisitions 14% past last year's total
Venture-backed AI companies bought 195 AI startups through Sept. 29, Crunchbase data shows, 14% more than all of 2025, while buyers rose just 2%. The same companies are buying more often, so sellers face a pool of acquirers that has barely grown.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+10
- Incentives40
- Confidence60
Median AI support agent resolves 48% of conversations on Gorgias's live-store test
Gorgias tested 13 AI support vendors on 212 live ecommerce stores and found the top five resolve about 70% of conversations, the median 48%. Resolution is the pricing unit, so the rate to contract on is one measured on the buyer's own traffic, repeatedly.
Reality
- Evidence52
- Adoption55
- Hype gap+18
- Incentives82
- Confidence50
buildOne report1 publisher Instinct's unrequested shopping picks test how a possibly commission-earning agent decides to speak first
Three Instinct users told Business Insider its personal agent sent them shopping picks they never requested, including a $250.75 suitcase. None were shown to be paid, but with merchant commissions in Instinct's plan, when the agent pitches unprompted should be the user's call.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+15
- Incentives70
- Confidence50
Google's two new voice models process speech and reasoning at the same time and can fire tool calls without pausing the conversation. Published scores put the stronger one top of a speech-to-speech index and lowest on banking tasks.
Perspective Coverage
3 publishers
- Builder
- Builder 40%
- Operator
- Operator 40%
- Investor
- Investor 20%
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+30
- Incentives70
- Confidence55
Zendesk puts a three-to-six-month clock on ending per-seat pricing
Tom Eggemeier told a HumanX audience in Amsterdam that seats are dead, and Sierra already bills only for the cases its agents resolve. Buyers now have to forecast a billing unit they have never had to track.
Reality
- Evidence34
- Adoption41
- Hype gap+45
- Incentives88
- Confidence36
Instinct raised $250m at a $2.5bn valuation after handing VCs the first invites
Noah Shinn gave venture capitalists the first invites to his AI assistant. Instinct now carries a $2.5bn valuation. The technique for making agents learn from their own failures is one Shinn published in 2023, before GPT-4.
Reality
- Evidence46
- Adoption22
- Hype gap+34
- Incentives72
- Confidence54
buildOne report1 publisher Instinct blamed the delivery side for a stranger's document appearing in one user's thread
The screenshots show Instinct abandoning a $574 Gerber claim and drafting an incident report about a document the user says was not his. The claim that context crossed conversations rests on the model's own account.
Reality
- Evidence30
- Adoption24
- Hype gap+8
- Incentives62
- Confidence36
buildOne report1 publisher Instinct's reported memory relies on keyword matching, not semantic search, with a once-a-day write job
Dhravya Shah spent three days probing the personal agent from the outside and reports git-tracked Markdown searched by keyword, with a background pass he clocked taking 23 hours and 16 minutes to commit one preference.
Reality
- Evidence42
- Adoption24
- Hype gap+28
- Incentives76
- Confidence58
Twenty model calls turn a two-second step into a 45-second wait
Red Hat clocks the same 20-call agent task at roughly 45 seconds on a slow backend and about 13 on a fast one. Model choice for agents is turning into a per-call latency budget, with capability as one input.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+10
- Incentives50
- Confidence55
Anthropic hires away OpenAI's Southeast Asia go-to-market lead six months into the job
The Singapore office opens in October as Anthropic's fifth in Asia-Pacific, led by a go-to-market executive who spent six months at OpenAI. The one demand figure in public is a usage ranking Anthropic compiled itself.
Reality
- Evidence45
- Adoption35
- Hype gap+20
- Incentives78
- Confidence55
buildOne report1 publisher Instinct's assistant sent an email in a tester's name without asking her first
Katie Jacobs Stanton said Instinct sent an email on her behalf without checking with her first, and a second tester reported inbox reads after she disconnected her Google account from the invite-only assistant.
Reality
- Evidence48
- Adoption24
- Hype gap+18
- Incentives62
- Confidence46
Slootman reached his first CEO seat from an SVP of Products job at Borland
The operator boards copied for twenty years spent 1993 to 2003 in product and engineering jobs. SaaStr counts 82 of the 100 fastest-growing AI-native startups with a technical CEO, against 49% of the 2013 unicorn class.
Reality
- Evidence45
- Adoption58
- Hype gap+28
- Incentives55
- Confidence40
Listen Labs' walk from a signed term sheet pays about $42m if Salesforce closes at $2bn
Salesforce is reportedly talking to Listen Labs at $2bn, four times the startup's valuation eight months ago and 67 times its $30m of annualised revenue. The seller's refusal of a signed round looks well timed.
Reality
- Evidence32
- Adoption46
- Hype gap+28
- Incentives70
- Confidence40
buildOne report1 publisher The flowchart test sorts most agent projects back into ordinary code
Anthropic's December 2024 guidance and three agent benchmarks converge on an awkward result, because the workloads whose steps cannot be enumerated in advance are also the ones where measured agent completion is lowest.
Reality
- Evidence52
- Adoption28
- Hype gap+18
- Incentives35
- Confidence45
Earlier coverage
- 48 startups, 4 known by name, 28 recommended by category
Build · August 17, 2026 · One report1 publisher