noiseorsignal.exetoday

NoiseOrSignal

./scorecardgraded: 0

Scorecard

Each call makes dated predictions at 12, 24 and 36 months. When one comes due, it is graded HIT or MISS here, in public.

n/aHit rate
93Predictions
0Graded
0Calls graded
28 Sep 2027Next due
HorizonPredictionsHitsMissesHit rate
12m31 (31 pending)00n/a
24m31 (31 pending)00n/a
36m31 (31 pending)00n/a
./pending93

PENDING

CallPredictionHorizonDueStatus
#1 OpenAI agent patternMore public incident reports.12mPENDING
#2 OpenAI training pauseSandboxing and agent monitoring become standard budget lines.12mPENDING
#3 Nvidia agent-containment platformSandbox runtimes could become a default part of agent deployments.12mPENDING
#5 Google Gemini 4 ArgonRestricted rollouts become the usual approach for the strongest models at the major labs.12mPENDING
#6 Cloudflare ClefDecision models become a normal step in automated workflows.12mPENDING
#7 OpenAI safety researchersMore labs tighten their rules on sharing internal information.12mPENDING
#8 KolibriMore models focus on a specific language and local requirements.12mPENDING
#9 125B model on one gaming GPULocal 100B-class models become normal for dev tools.12mPENDING
#10 Claude chat led to arrestMore AI firms publish clear reporting rules.12mPENDING
#11 EmbeddingGemma 2I expect more apps to offer local search.12mPENDING
#12 OpenSEOI expect AI visibility reports in most SEO tools.12mPENDING
#13 Whistle speech-to-textI expect more apps to do basic transcription on the device.12mPENDING
#14 Decision modelsDecision APIs become the default for structured tasks.12mPENDING
#15 Anthropic S-1 IPOAnthropic’s quarterly reports become the AI sector’s reality check.12mPENDING
#16 Meta Muse WonderMeta stays a serious API option for agents.12mPENDING
#17 Anthropic source controlMore teams will hit review and CI limits as agents write more code.12mPENDING
#18 NVIDIA Run:ai open sourceMore GPU platforms will pick up parts of Run:ai.12mPENDING
#19 Meta Muse Spark OpenOpen agentic models become the default for internal dev tools.12mPENDING
#20 ECDSA curve originsModels routinely assist on open research problems.12mPENDING
#21 DeepSeek V4.1 FlashOpen models take most agentic coding volume on cost.12mPENDING
#25 LiteLLM PyPI compromiseAI libraries become the top supply-chain target.12mPENDING
#26 PHP bug found by AIAI-found CVEs are routine in major open-source projects.12mPENDING
#27 Kimi K3.1 ProOpen-weight coding agents are the default for cost-sensitive teams.12mPENDING
#28 No AI payoff outside techBoards demand hard ROI numbers for AI budgets.12mPENDING
#29 Small CNN beats LLMs on skin lesionsDomain teams run specialist models beside one general LLM.12mPENDING
#30 OpenAI claims three open-problem proofsAI-assisted proofs are routine in theory work.12mPENDING
#32 Gemini 4 Carbon gated to defendersMore frontier launches start with restricted defender access.12mPENDING
#33 Mistral Large 4 open weightsMore teams evaluate open-weight models for specific coding workflows.12mPENDING
#34 Cursor Claude Code extension backdoorExtension provenance gets real scrutiny.12mPENDING
#35 ChatGPT shared-link chat leakShare and preview surfaces get login-level security review.12mPENDING
#36 China AI model export rulesMore model evaluations include deployment and licensing constraints.12mPENDING
#1 OpenAI agent patternTighter controls become a buying requirement.24mPENDING
#2 OpenAI training pauseContainment audits become a procurement requirement, like SOC2.24mPENDING
#3 Nvidia agent-containment platformHardware-based monitoring could show up in procurement requirements.24mPENDING
#5 Google Gemini 4 ArgonAccess to those models becomes part of enterprise contract negotiations.24mPENDING
#6 Cloudflare ClefMost routing calls no longer need a large general-purpose LLM.24mPENDING
#7 OpenAI safety researchersMore buyers ask for independent safety audits before signing contracts.24mPENDING
#8 KolibriMore buyers ask where and how a model was trained.24mPENDING
#9 125B model on one gaming GPUPrivate coding agents run on laptops.24mPENDING
#10 Claude chat led to arrestEnterprise buyers demand zero-retention terms.24mPENDING
#11 EmbeddingGemma 2I expect search across media to become common.24mPENDING
#12 OpenSEOI expect teams to track mentions and citations, not only rankings.24mPENDING
#13 Whistle speech-to-textI expect voice to show up as a normal input in more tools.24mPENDING
#14 Decision modelsEvery major cloud ships a decision model family. Prices keep falling.24mPENDING
#15 Anthropic S-1 IPOEvery frontier lab faces the same pressure to show real numbers.24mPENDING
#16 Meta Muse WonderMeta’s pricing puts pressure on everyone else to lower agent costs.24mPENDING
#17 Anthropic source controlDeveloper tools will compete on safe throughput, not just how much code they can produce.24mPENDING
#18 NVIDIA Run:ai open sourceBasic scheduling features will be harder to sell on their own.24mPENDING
#19 Meta Muse Spark OpenDesktop-class frontier models put pressure on API pricing.24mPENDING
#20 ECDSA curve originsAI-assisted discoveries become normal in security audits.24mPENDING
#21 DeepSeek V4.1 FlashSearch and fetch become metered utilities inside agent stacks.24mPENDING
#25 LiteLLM PyPI compromiseSigned, reproducible builds are table stakes for AI packages.24mPENDING
#26 PHP bug found by AICoordinated disclosure with AI labs is standard practice.24mPENDING
#27 Kimi K3.1 ProModel choice is a routing decision.24mPENDING
#28 No AI payoff outside techThe winners are vertical tools with proven unit economics.24mPENDING
#29 Small CNN beats LLMs on skin lesionsRegulated fields default to auditable task models, with LLMs writing around them.24mPENDING
#30 OpenAI claims three open-problem proofsFormal verification is standard for major claims.24mPENDING
#32 Gemini 4 Carbon gated to defendersCyber-focused model offerings become a clear enterprise category.24mPENDING
#33 Mistral Large 4 open weightsDeployment control becomes a stronger buying criterion.24mPENDING
#34 Cursor Claude Code extension backdoorDevelopers demand narrower permissions for editor extensions.24mPENDING
#35 ChatGPT shared-link chat leakCross-frame message validation shows up in AI-product audit checklists.24mPENDING
#36 China AI model export rulesExport terms influence more hosting decisions.24mPENDING
#1 OpenAI agent patternAgent trust becomes part of product quality, not a footnote.36mPENDING
#2 OpenAI training pauseLiability for autonomous agent actions lands in contracts and courts.36mPENDING
#3 Nvidia agent-containment platformInsurers could start asking how agents are contained and pricing coverage accordingly.36mPENDING
#5 Google Gemini 4 ArgonSome access restrictions become regulatory requirements.36mPENDING
#6 Cloudflare ClefThe big labs offer their own decision models.36mPENDING
#7 OpenAI safety researchersFormal reporting requirements start to determine which safety incidents labs must disclose.36mPENDING
#8 KolibriMore regions have models built for their regulated industries.36mPENDING
#9 125B model on one gaming GPUCloud APIs keep only the hardest jobs.36mPENDING
#10 Claude chat led to arrestLaws decide what AI firms must report.36mPENDING
#11 EmbeddingGemma 2I expect buyers to demand clear choices about where indexing happens.36mPENDING
#12 OpenSEOI expect paid SEO suites to compete with open tools on price.36mPENDING
#13 Whistle speech-to-textI expect cloud speech APIs to handle only the hard cases.36mPENDING
#14 Decision modelsAgent systems split into planners that chat and deciders that score.36mPENDING
#15 Anthropic S-1 IPOAI valuations track audited unit economics more than demos.36mPENDING
#16 Meta Muse WonderModel choice turns into a routing decision instead of a long-term vendor bet.36mPENDING
#17 Anthropic source controlTests and permissions will matter more than the volume of code agents can write.36mPENDING
#18 NVIDIA Run:ai open sourceSupport and reliable operations will matter more than whether the scheduler is open.36mPENDING
#19 Meta Muse Spark OpenThe open-weight gap between countries matters more than the closed-model race.36mPENDING
#20 ECDSA curve origins“AI found it” becomes a standard line in papers.36mPENDING
#21 DeepSeek V4.1 FlashModel choice matters less than tool economics.36mPENDING
#25 LiteLLM PyPI compromiseTeams isolate agent runtimes from credential stores by default.36mPENDING
#26 PHP bug found by AIShipping code without AI security review looks negligent.36mPENDING
#27 Kimi K3.1 ProFrontier capability concentrates in a few labs and the rest is commoditised.36mPENDING
#28 No AI payoff outside techThe payoff finally shows in the data, built on boring workflow wins from this era.36mPENDING
#29 Small CNN beats LLMs on skin lesions“Which model” is a per-task routing call made on private evals.36mPENDING
#30 OpenAI claims three open-problem proofsResearch splits into generation, verification, and explanation.36mPENDING
#32 Gemini 4 Carbon gated to defendersAccess rules draw regulatory scrutiny.36mPENDING
#33 Mistral Large 4 open weightsHosted and self-hosted models run side by side in most stacks.36mPENDING
#34 Cursor Claude Code extension backdoorAI-tool supply chains are routine security-review material.36mPENDING
#35 ChatGPT shared-link chat leakTesting playbooks cover exploit chains that span AI content and browser features.36mPENDING
#36 China AI model export rulesMore products keep region-specific model options.36mPENDING
./hit0

HIT

No hits graded yet. First predictions come due 28 Sep 2027.

./miss0

MISS

No misses graded yet.