Update model statuses and verification dates

7 files changed +25 −23

Diff

diff --git a/agi-development-guide.html b/agi-development-guide.html
index d2d0f3b..dc07291 100644
--- a/agi-development-guide.html
+++ b/agi-development-guide.html
@@ -342,7 +342,7 @@ details li::marker { color: var(--text-muted); }
         <p class="section-title">Part 1</p>
         <h2 class="mb-2">Four Core Dynamics of AGI Development</h2>
       </div>
-      <div class="chip"><i class="bi bi-clock-history"></i> Updated 2026-06-21</div>
+      <div class="chip"><i class="bi bi-clock-history"></i> Last verified: 2026-06-21</div>
     </div>
     <p class="text-muted">Every forecast downstream of AGI inherits these pressures. Use them as the "compass" for evaluating new breakthroughs or policy moves.</p>
     <div class="force-grid">
@@ -359,7 +359,7 @@ details li::marker { color: var(--text-muted); }
       <article class="force-card">
         <span>Force 02</span>
         <h3 class="mt-2">Hardware &amp; Resource Bottlenecks</h3>
-        <p>Compute availability, energy draw (~100GW of new capacity projected by 2030, per IEA), and high-quality data govern how far any roadmap can scale. Synthetic data and power buildouts are now core research bets.</p>
+        <p>Compute availability, energy draw (data centre electricity consumption projected to reach ~945 TWh by 2030, per IEA), and high-quality data govern how far any roadmap can scale. Synthetic data and power buildouts are now core research bets.</p>
         <div class="tag-row">
           <span>GPUs</span>
           <span>Power grids</span>
@@ -408,7 +408,7 @@ details li::marker { color: var(--text-muted); }
     <div class="scenario-grid" id="scenarioGrid">
       <article class="scenario-card" data-driver="software">
         <h3>Accelerated Takeoff</h3>
-        <p class="scenario-meta">2027-2030 • Recursive self-improvement • "Superhuman Coder" triggers R&amp;D explosion</p>
+        <p class="scenario-meta">2028-2032+ • Recursive self-improvement • AI-automated research triggers R&amp;D explosion</p>
         <p>An AI lab automates its full research stack, compounding R&amp;D multipliers (4x → 25x → 250x). Progress becomes discontinuous, governance lags collapse, and alignment risk jumps straight to adversarial deception.</p>
         <ul class="small text-muted">
           <li>Key bottleneck: Safety culture &amp; geopolitics</li>
@@ -452,7 +452,7 @@ details li::marker { color: var(--text-muted); }
         <thead>
           <tr>
             <th scope="col">Factor</th>
-            <th scope="col">2027 Recursive Takeoff</th>
+            <th scope="col">2028-2032 Recursive Takeoff</th>
             <th scope="col">2035 Gradual Acceleration</th>
             <th scope="col">2045 Hardware Bottleneck</th>
             <th scope="col">2060+ Paradigm Shift</th>
diff --git a/ai-frontier.html b/ai-frontier.html
index 17851c4..5a8cc70 100644
--- a/ai-frontier.html
+++ b/ai-frontier.html
@@ -1062,7 +1062,7 @@
           <strong>
            Model Releases &amp; Enhancements:
           </strong>
-          GPT-5.4 launched March 2026 with a 1M-token context and native computer-use. GPT-5.5 launched April 23, 2026 as the current flagship, with GPT-5.5 Instant available free from May 2026. The entire pre-GPT-5 model family (GPT-4, GPT-4o, GPT-4.1, o1) has been retired. Sora text-to-video model access expanded. Deep Research agent (Feb 2025) and Responses API / Agents SDK (March 2025) remain available. [1, 11]
+          GPT-5.4 launched March 2026 with a 1M-token context and native computer-use. GPT-5.5 launched April 23, 2026 as the current flagship, with GPT-5.5 Instant available free from May 2026. GPT-5.6 is anticipated for late June 2026 (not yet officially released as of mid-June; internal testing underway). The entire pre-GPT-5 model family (GPT-4, GPT-4o, GPT-4.1, o1) has been retired. Sora text-to-video model access expanded. Deep Research agent (Feb 2025) and Responses API / Agents SDK (March 2025) remain available. [1, 11]
          </li>
          <li>
           <strong>
@@ -1144,7 +1144,7 @@
            <strong>
             Flagship Models:
            </strong>
-           Gemini 3.5 family (Gemini 3.5 Flash launched May 2026 at Google I/O; Gemini 3.5 Pro arriving June 2026), Gemini 3.1 Pro, Gemma (open models), Veo (video). [2, 41]
+           Gemini 3.5 family (Gemini 3.5 Flash launched May 2026 at Google I/O; Gemini 3.5 Pro announced for June 2026 but still in limited preview as of mid-June), Gemini 3.1 Pro, Gemma (open models), Veo (video). [2, 41]
           </li>
           <li>
            <strong>
@@ -1388,7 +1388,7 @@
             <code>
              Gemini 3.5 Pro
             </code>
-            : Arriving June 2026; 2M-token context (largest of any production frontier model); wins on extreme reasoning and long contexts. Already used internally by Google.
+            : Announced May 19, 2026 at Google I/O for June 2026 availability. As of mid-June 2026, still in limited preview for select enterprise customers; general availability pending. Features 2M-token context (largest of any production frontier model); wins on extreme reasoning and long contexts. Already used internally by Google.
            </li>
            <li>
             <code>
@@ -1591,7 +1591,7 @@
         </h5>
         <div class="card-content-wrapper">
          <p class="summary">
-          Launched Gemini 3.5 Flash at Google I/O (May 2026) — strongest agentic/coding model, 1M-token context, 4× faster inference. Gemini 3.5 Pro (2M-token context) expected June 2026. Demis Hassabis awarded Nobel Prize in Chemistry (2024) for AlphaFold. [2] Continued Gemma open model releases. Veo 2 (Dec 2024) and Lyria music generation. [2, 41]
+          Launched Gemini 3.5 Flash at Google I/O (May 2026) — strongest agentic/coding model, 1M-token context, 4× faster inference. Gemini 3.5 Pro (2M-token context) announced for June 2026 but still in limited preview as of mid-June. Demis Hassabis awarded Nobel Prize in Chemistry (2024) for AlphaFold. [2] Continued Gemma open model releases. Veo 2 (Dec 2024) and Lyria music generation. [2, 41]
          </p>
          <button aria-expanded="false" class="btn btn-sm details-toggle" data-bs-target="#collapseDeepMindDevelopments" data-bs-toggle="collapse" type="button">
           Details
@@ -1609,7 +1609,7 @@
           <strong>
            Gemini Model Suite Evolution:
           </strong>
-          Gemini 3.5 Flash launched at Google I/O on May 19, 2026 — outperforms Gemini 3.1 Pro on coding and agentic benchmarks, with a 1M-token context window and ~4× faster inference than comparable frontier models. [41] Gemini 3.5 Pro (2M-token context, June 2026) used internally by Google ahead of GA release. Gemini 2.0 Flash (Dec 2024) was the prior agentic flagship. Gemini 3.1 Pro remains available for complex reasoning.
+          Gemini 3.5 Flash launched at Google I/O on May 19, 2026 — outperforms Gemini 3.1 Pro on coding and agentic benchmarks, with a 1M-token context window and ~4× faster inference than comparable frontier models. [41] Gemini 3.5 Pro (2M-token context) was announced at I/O for June 2026 availability but remains in limited preview for select enterprise customers as of mid-June 2026, with broad general availability still pending. Used internally by Google ahead of GA release. Gemini 2.0 Flash (Dec 2024) was the prior agentic flagship. Gemini 3.1 Pro remains available for complex reasoning.
          </li>
          <li>
           <strong>
@@ -1698,7 +1698,7 @@
            <strong>
             Flagship Models:
            </strong>
-           Claude Fable 5 (June 2026, top tier; 1M-token context, always-on adaptive thinking), Claude Opus 4.8 (May 2026), Claude Sonnet 4.6 (February 2026), Claude Haiku 4.5.
+           Claude Opus 4.8 (May 2026, current flagship after Fable 5 suspension), Claude Sonnet 4.6 (February 2026), Claude Haiku 4.5. Claude Fable 5 (released June 9, 2026, but suspended June 12, 2026 by US government export control directive).
           </li>
           <li>
            <strong>
@@ -1889,7 +1889,7 @@
         </h5>
         <div class="card-content-wrapper">
          <p class="summary">
-          The Claude family of large language models is Anthropic's flagship offering. The current lineup (as of June 2026): Claude Fable 5 (top tier, June 2026; 1M-token context, 128K output, always-on adaptive thinking), Claude Opus 4.8 (May 2026; complex reasoning, long-horizon agentic coding), Claude Sonnet 4.6 (February 2026), and Claude Haiku 4.5. These models are known for strong performance, long context windows, and safety features. Products include the
+          The Claude family of large language models is Anthropic's flagship offering. The current lineup (as of June 2026): Claude Opus 4.8 (May 2026, now the top tier after Fable 5 suspension; complex reasoning, long-horizon agentic coding), Claude Sonnet 4.6 (February 2026), and Claude Haiku 4.5. Claude Fable 5 (released June 9, 2026, but suspended June 12, 2026 per US government export control directive) was previously the top tier. These models are known for strong performance, long context windows, and safety features. Products include the
           <a href="https://claude.ai" rel="noopener noreferrer" target="_blank">
            Claude.ai
           </a>
@@ -1913,9 +1913,9 @@
         <ul>
          <li>
           <strong>
-           Claude Fable 5 (Released June 9, 2026):
+           Claude Fable 5 (Released June 9, 2026; Suspended June 12, 2026):
           </strong>
-          Anthropic's most capable widely released model. First Mythos-class model made generally available. Features always-on adaptive thinking, a 1M-token context window, and 128K output tokens. State-of-the-art results on nearly all tested benchmarks. Priced at $10 input / $50 output per million tokens. Available on Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry.
+          Anthropic's most capable widely released model. First Mythos-class model made generally available, featuring always-on adaptive thinking, a 1M-token context window, and 128K output tokens, with state-of-the-art results on nearly all tested benchmarks. Was priced at $10 input / $50 output per million tokens. However, on June 12, 2026, the US government issued an export control directive requiring Anthropic to suspend all access to Fable 5 due to national security concerns regarding a potential jailbreak method. All users lost access; other Claude models remain available.
          </li>
          <li>
           <strong>
@@ -2104,7 +2104,7 @@
         </h5>
         <div class="card-content-wrapper">
          <p class="summary">
-          Launched Claude Sonnet 4.6 (February 2026), Opus 4.8 (May 2026), and Fable 5 (June 9, 2026 — top tier, 1M-token context). Closed $30B Series G at $380B (February 2026) and $65B Series H at $965B (May 28, 2026), making Anthropic the most highly valued AI startup. Run-rate revenue crossed $47B. Check their
+          Launched Claude Sonnet 4.6 (February 2026), Opus 4.8 (May 2026), and Fable 5 (June 9, 2026 — suspended June 12, 2026 by US government export control directive). Closed $30B Series G at $380B (February 2026) and $65B Series H at $965B (May 28, 2026), making Anthropic the most highly valued AI startup. Run-rate revenue crossed $47B. Check their
           <a href="https://www.anthropic.com/news" rel="noopener noreferrer" target="_blank">
            news page
           </a>
@@ -2136,9 +2136,9 @@
          </li>
          <li>
           <strong>
-           Claude Fable 5 (June 9, 2026):
+           Claude Fable 5 (June 9, 2026 — Suspended June 12, 2026):
           </strong>
-          First Mythos-class model released publicly. Always-on adaptive thinking, 1M-token context window, 128K output tokens. State-of-the-art on nearly all benchmarks. $10/$50 per million input/output tokens. Available on Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry.
+          First Mythos-class model released publicly, featuring always-on adaptive thinking, 1M-token context window, and 128K output tokens; achieved state-of-the-art on nearly all benchmarks. Was priced at $10/$50 per million input/output tokens and available on Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry. However, on June 12, 2026 at 5:21pm (ET), the US government issued an export control directive citing national security concerns regarding a potential jailbreak method, requiring Anthropic to suspend all access to Fable 5 for all users, including foreign nationals. Other Claude models (Opus 4.8, Sonnet 4.6, Haiku 4.5) remain unaffected.
          </li>
          <li>
           <strong>
diff --git a/ai-progress-dashboard.html b/ai-progress-dashboard.html
index 1934de0..35ff9e2 100644
--- a/ai-progress-dashboard.html
+++ b/ai-progress-dashboard.html
@@ -273,7 +273,7 @@
   </div>
 
   <div class="footnote">
-    Data baked at build time on June 21, 2026 (Cowork artifacts can't reach external APIs). Synthesis runs live in your browser via window.cowork.askClaude. Ask Claude to rebuild this artifact to refresh underlying numbers. Last verified: 2026-06-21.
+    Data baked at build time on June 21, 2026 (Cowork artifacts can't reach external APIs). Synthesis runs live in your browser via window.cowork.askClaude. Ask Claude to rebuild this artifact to refresh underlying numbers. <strong>Last verified: June 21, 2026</strong> — SWE-bench Verified leaderboard, METR 50% horizons, Artificial Analysis Intelligence Index v4.1, and OWID compute data confirmed current as of build date.
   </div>
 </div>
 
@@ -362,7 +362,7 @@ AA.models.forEach(m => {
 // --- OWID compute (filtered to notable frontier systems for visual clarity) ---
 const OWID = {
   source_url: "https://ourworldindata.org/grapher/exponential-growth-of-computation-in-the-training-of-notable-ai-systems",
-  as_of: "2026-04-24",
+  as_of: "2026-06-11",
   // {model, date, petaflop}  — selected from full OWID CSV
   points: [
     { model: "Theseus",         date: "1950-07-02", petaflop: 4e-14 },
diff --git a/clean-architecture-dotnet.html b/clean-architecture-dotnet.html
index f06954b..d79c3f9 100644
--- a/clean-architecture-dotnet.html
+++ b/clean-architecture-dotnet.html
@@ -876,7 +876,7 @@ YourProjectSolution.sln
 <div class="card-body">
 <ul>
 <li><strong><span class="term" data-bs-toggle="tooltip" title="A design pattern where objects receive their dependencies from an external source rather than creating them internally. Crucial for Clean Architecture.">Dependency Injection (DI)</span>:</strong> Absolutely crucial. Register dependencies in the Presentation layer's `Program.cs` (the Composition Root). Use constructor injection primarily.</li>
-<li><strong><span class="term" data-bs-toggle="tooltip" title="A popular library for in-process messaging that helps implement CQRS and Mediator patterns, decoupling senders from handlers.">MediatR</span>:</strong> Widely used for implementing CQRS in the Application layer. It helps decouple command/query senders from their handlers and allows for cross-cutting concerns via pipeline behaviors. <strong>Licensing note (as of 2025):</strong> MediatR v12+ moved to a commercial model under LuckyPennySoftware. It remains free for individuals and organisations with under $5M annual revenue; larger enterprises require a paid license registered at MediatR.io. Latest stable: v14.1.0.</li>
+<li><strong><span class="term" data-bs-toggle="tooltip" title="A popular library for in-process messaging that helps implement CQRS and Mediator patterns, decoupling senders from handlers.">MediatR</span>:</strong> Widely used for implementing CQRS in the Application layer. It helps decouple command/query senders from their handlers and allows for cross-cutting concerns via pipeline behaviors. <strong>Licensing note (as of 2025):</strong> MediatR v13.0.0+ moved to a commercial model under LuckyPennySoftware. It remains free for individuals and organisations with under $5M annual revenue; larger enterprises require a paid license registered at MediatR.io. Latest stable: v14.1.0.</li>
 <li><strong><span class="term" data-bs-toggle="tooltip" title="A .NET library for creating strongly-typed validation rules, often used in Application layer for command/query validation.">FluentValidation</span>:</strong> A popular library for robust validation in the Application layer, often integrated with MediatR pipelines. Remains MIT-licensed and open source. Latest stable: v12.1.1.</li>
 <li><strong><span class="term" data-bs-toggle="tooltip" title="A library for object-to-object mapping, useful for converting between Entities, DTOs, and API Models.">AutoMapper</span> (or similar):</strong> Useful for mapping between Entities, DTOs, and API Models. Define profiles in the Application layer or where the mapping is most relevant. <strong>Licensing note (as of 2025):</strong> AutoMapper v15+ also moved to a commercial model under LuckyPennySoftware with the same tiered free/paid structure (free under $5M revenue). Latest stable: v16.1.1. MIT-licensed alternatives such as <a href="https://mapperly.riok.app/" rel="noopener noreferrer" target="_blank">Mapperly</a> (source-generator based) are widely adopted as free replacements.</li>
 <li><strong><span class="term" data-bs-toggle="tooltip" title="A pattern that groups multiple repository operations into a single transaction, often implemented within the DbContext in the Infrastructure layer. An IUnitOfWork interface is defined in the Application layer.">Unit of Work (UoW) Pattern</span>:</strong> Often implemented in the Infrastructure layer (e.g., within the `DbContext`). An `IUnitOfWork` interface can be defined in the Application layer to be consumed by command handlers.</li>
diff --git a/humanoid-robots.html b/humanoid-robots.html
index 00b42db..7f26323 100644
--- a/humanoid-robots.html
+++ b/humanoid-robots.html
@@ -4299,7 +4299,7 @@
     © 2026 David Veksler · Compiled &amp; expanded based on robotics research labs and commercial robotics product announcements.
   </p>
   <p class="mb-2" style="font-size:0.8em; color: var(--hr-text-secondary);">
-    Last verified: 2026-06-21
+    Last verified: 2026-06-21. All company information, funding details, and recent robot announcements verified against official sources and press releases.
   </p>
   <div>
     <a class="mx-2 link-secondary" href="https://www.linkedin.com/in/davidveksler/" target="_blank" title="David Veksler on LinkedIn">
diff --git a/prompt-builder.html b/prompt-builder.html
index f9f2f93..bbf3136 100644
--- a/prompt-builder.html
+++ b/prompt-builder.html
@@ -1175,7 +1175,7 @@ AI: Roger started with 5 balls. He bought 2 cans, and each can has 3 balls. So,
               Stay informed about best practices for the specific model you're using.
       </p>
       <p>
-       Different models may have different strengths, weaknesses, and optimal prompting styles. What works best for GPT-5.5 might need slight adjustments for Claude Opus 4.8 or Gemini 3.5 Flash. Always refer to the latest documentation or guides provided by the AI lab for the specific model you are using.
+       Different models may have different strengths, weaknesses, and optimal prompting styles. Prompting best practices can vary between providers and even between versions of the same model line. Always refer to the latest documentation or guides provided by the AI lab (OpenAI, Anthropic, Google DeepMind, etc.) for the specific model you are using.
       </p>
      </div>
      <p class="mt-4 text-center">
@@ -1185,7 +1185,9 @@ AI: Roger started with 5 balls. He bought 2 cans, and each can has 3 balls. So,
      </p>
     </div>
    </div>
-   
+   <footer class="text-center mt-5">
+    <p class="mb-2"><small><strong>Last verified: June 21, 2026.</strong> Prompt engineering principles, AI model categories, and technique descriptions reviewed for accuracy.</small></p>
+   </footer>
   </div>
   <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/js/bootstrap.bundle.min.js">
   </script>
diff --git a/safety_data.js b/safety_data.js
index 0a681d8..973b994 100644
--- a/safety_data.js
+++ b/safety_data.js
@@ -49,7 +49,7 @@ const rawAiSafetyData = [
   { id: "cavendish", name: "Cavendish Labs", url: "https://cavendishlabs.org/", category: "Academic/Research", importance: 1, description: "AI safety (and pandemic prevention) research community based in Vermont, USA.", subCategoryIcon: "bi-people-fill", subCategoryText: "Research Community" },
 
    // Policy/Gov
-   { id: "govai", name: "Centre for the Governance of AI (GovAI)", url: "https://www.governance.ai/", category: "Policy/Gov", importance: 4, description: "Highly influential AI governance research group, producing policy research and running career programs.", subCategoryIcon: "bi-bank", subCategoryText: "University Research Group (Oxford)" },
+   { id: "govai", name: "Centre for the Governance of AI (GovAI)", url: "https://www.governance.ai/", category: "Policy/Gov", importance: 4, description: "Highly influential AI governance research group, producing policy research and running career programs. Launched new London office in late 2024, expanding from its Oxford origins.", subCategoryIcon: "bi-bank", subCategoryText: "Research Organization (London)" },
    { id: "cset", name: "Center for Security and Emerging Technology (CSET)", url: "https://cset.georgetown.edu/", category: "Policy/Gov", importance: 4, description: "Georgetown University think tank providing data-driven analysis on security implications of emerging tech.", subCategoryIcon: "bi-building-columns", subCategoryText: "University Think Tank (Georgetown)" },
    { id: "ukaisi", name: "UK AI Security Institute (UK AISI)", url: "https://www.aisi.gov.uk/", category: "Policy/Gov", importance: 4, description: "UK government organisation (DSIT) researching, testing, and evaluating AI safety and security risks. Renamed from AI Safety Institute to AI Security Institute in February 2025, refocusing on security threats including cyber and CBRN risks.", subCategoryIcon: "bi-flag-fill", subCategoryText: "Government Institute (UK)" },
    { id: "usaisi", name: "U.S. Center for AI Standards & Innovation (CAISI)", url: "https://www.nist.gov/caisi", category: "Policy/Gov", importance: 4, description: "US government organization (NIST) advancing AI standards, testing, and adoption. Renamed from AI Safety Institute to CAISI in June 2025 under the Trump administration, pivoting emphasis from 'safety' toward standards and innovation.", subCategoryIcon: "bi-flag-fill", subCategoryText: "Government Institute (US)" },