{"assertions":[{"assertion_index":0,"assertion_text":"Anthropic also expects a profitable third quarter, judged by the same internal model it applied when it called the second quarter a profit, and it counts 6,000 customers spending $100,000 or more a year Simon Willison's Weblog.","citation_check":null,"claim_ids":[50954,50955,50956],"link_only_urls":[]},{"assertion_index":1,"assertion_text":"The same reporting undercuts that comfort: it frames Anthropic's best model as struggling to attract users even as cheaper tools thrive, and Anthropic's rapid revenue growth has been running alongside cheaper open-source alternatives for a while Nate Jones.","citation_check":null,"claim_ids":[10156],"link_only_urls":[]},{"assertion_index":2,"assertion_text":"Z.ai's $18-a-month GLM plan, built on GLM-5.3, now carries official compatibility with both Claude Code and Codex, the coding agents from Anthropic and OpenAI, which effectively lets those tools serve as the working surface for a discount model Nate Jones.","citation_check":null,"claim_ids":[50787],"link_only_urls":[]},{"assertion_index":3,"assertion_text":"Nate Jones's cost scorecard projects net savings above $132 from moving a fraction of work to GLM-5.3 instead of running only the expensive models Nate Jones.","citation_check":null,"claim_ids":[50145,50148],"link_only_urls":[]},{"assertion_index":4,"assertion_text":"Investors expect Anthropic to finish 2026 between $100 billion and $120 billion in revenue, which only works if today's premium defaults hold TechCrunch AI.","citation_check":null,"claim_ids":[47997,47998,48000],"link_only_urls":[]},{"assertion_index":5,"assertion_text":"OpenAI has nineteen open requisitions for forward-deployed engineers, the people embedded with customers to make AI work inside real systems, paying $162,000 to $280,000 plus equity in San Francisco, and Handshake lists a senior version at up to $350,000 Nate Jones.","citation_check":null,"claim_ids":[50930,50931,50932],"link_only_urls":[]},{"assertion_index":6,"assertion_text":"market holds roughly 17,000 such engineers, many at Palantir TechCrunch AI.","citation_check":null,"claim_ids":[39855],"link_only_urls":[]},{"assertion_index":7,"assertion_text":"The answer from the EU AI Act Service Desk is no: the Act's existing definitions of AI systems and general-purpose models already cover agents, and the AI Office says analyzing model usage can count as post-market monitoring, the required watching of a product after launch EU AI Act Newsletter.","citation_check":null,"claim_ids":[50961,50962,50963],"link_only_urls":[]},{"assertion_index":8,"assertion_text":"Because agents can change behavior after deployment, legal commentary expects compliance to be harder for them in practice, through the Act's substantial-modification rules and its human-oversight requirements EU AI Act Newsletter.","citation_check":null,"claim_ids":[22671,22683,22684],"link_only_urls":[]},{"assertion_index":9,"assertion_text":"On the second, Judge Stephanos Bibas reached the opposite result last year, finding no fair use where Thomson Reuters content trained a rival legal platform TechCrunch AI.","citation_check":null,"claim_ids":[50920,50921,50922],"link_only_urls":[]},{"assertion_index":10,"assertion_text":"- Anthropic's third-quarter disclosure will show whether the profit claim survives a quarter of cheap-model routing. Simon Willison's Weblog","citation_check":null,"claim_ids":[50954,50955,50956],"link_only_urls":[]},{"assertion_index":11,"assertion_text":"- OpenAI's forward-deployed engineer requisitions are a proxy for where the bottleneck sits; growth past nineteen says deployment labor, not model quality. Nate Jones","citation_check":null,"claim_ids":[50930,50931,50932],"link_only_urls":[]},{"assertion_index":12,"assertion_text":"- Z.ai's next pricing or compatibility move matters most, since official support inside rival coding tools is the wedge. Nate Jones","citation_check":null,"claim_ids":[50787],"link_only_urls":[]},{"assertion_index":13,"assertion_text":"- The EU AI Office's follow-up guidance on how existing definitions bind agent providers will set the compliance floor for anyone shipping agents in Europe. EU AI Act Newsletter","citation_check":null,"claim_ids":[50961,50962,50963],"link_only_urls":[]},{"assertion_index":14,"assertion_text":"- Appeals in the Thomson Reuters line of cases could harden the competing-product test that now splits the copyright picture. TechCrunch AI","citation_check":null,"claim_ids":[50920,50921,50922],"link_only_urls":[]},{"assertion_index":15,"assertion_text":"The $65 billion figure and the report that Anthropic's best model struggles to attract users rest on Financial Times reporting via unnamed people with knowledge of the matter, and Anthropic's profitability claims rest on its own internal financial model rather than audited results Simon Willison's Weblog.","citation_check":null,"claim_ids":[50954,50955,50956],"link_only_urls":[]}],"brief_date":"2026-08-24","bundle_id":"evidence-bundle-v1-e92bc937ed72b0728818de4cf5190bad796a68855d01c1de8b37cf6b5532e8fb","claims":[{"assertion_indexes":[0,10,15],"citation_authority_claim_id":50954,"citation_checks":[],"claim_id":50954,"claim_text":"Anthropic's annualized revenue for July 2026 was reported as up to $65 billion, compared to $47 billion in May 2026. (Source: FT story via people with knowledge of the matter)","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:35:04.871380+00:00","source_name":"Simon Willison's Weblog","source_url":"https://simonwillison.net/2026/Aug/23/anthropics-best-ai-model-struggles-to-attract-users-as-cheaper-t/"},{"assertion_indexes":[0,10,15],"citation_authority_claim_id":50955,"citation_checks":[],"claim_id":50955,"claim_text":"Anthropic expects its third quarter of 2026 to be profitable, using the same financial model it used to declare the second quarter profitable.","correction_role":null,"label":"forecast","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:35:04.871380+00:00","source_name":"Simon Willison's Weblog","source_url":"https://simonwillison.net/2026/Aug/23/anthropics-best-ai-model-struggles-to-attract-users-as-cheaper-t/"},{"assertion_indexes":[0,10,15],"citation_authority_claim_id":50956,"citation_checks":[],"claim_id":50956,"claim_text":"Anthropic stated it had 6,000 customers spending $100,000 or more annually.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:35:04.871380+00:00","source_name":"Simon Willison's Weblog","source_url":"https://simonwillison.net/2026/Aug/23/anthropics-best-ai-model-struggles-to-attract-users-as-cheaper-t/"},{"assertion_indexes":[1],"citation_authority_claim_id":10156,"citation_checks":[],"claim_id":10156,"claim_text":"Anthropic's revenue is currently growing rapidly despite the availability of cheaper open-source alternatives.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-09T06:21:54.201981+00:00","source_name":"Nate Jones","source_url":"https://www.youtube.com/watch?v=Zp8lr6IzUnQ"},{"assertion_indexes":[2,12],"citation_authority_claim_id":50787,"citation_checks":[],"claim_id":50787,"claim_text":"Z.ai's GLM coding plan costs $18 per month and is officially compatible with both Claude Code and Codex.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:30:56.319700+00:00","source_name":"Nate Jones","source_url":"https://www.youtube.com/watch?v=4HvFqhtCb-A"},{"assertion_indexes":[3],"citation_authority_claim_id":50145,"citation_checks":[],"claim_id":50145,"claim_text":"Z.AI's GLM Coding Plan costs $18 per month and is based on GLM-5.3.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-22T06:33:45.069663+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/glm-5-3-claude-code-codex"},{"assertion_indexes":[3],"citation_authority_claim_id":50148,"citation_checks":[],"claim_id":50148,"claim_text":"Moving a fraction of work to GLM-5.3 can result in net savings exceeding $132 compared to using only expensive models like Codex or Claude Code.","correction_role":null,"label":"forecast","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-22T06:33:45.069663+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/glm-5-3-claude-code-codex"},{"assertion_indexes":[4],"citation_authority_claim_id":47997,"citation_checks":[],"claim_id":47997,"claim_text":"Anthropic's annualized revenue run rate surpassed $65 billion at the end of July 2026, according to a Bloomberg report.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-18T07:35:18.103873+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/17/anthropics-annualized-revenue-surges-to-65b/"},{"assertion_indexes":[4],"citation_authority_claim_id":47998,"citation_checks":[],"claim_id":47998,"claim_text":"Anthropic's annualized revenue run rate was $47 billion in May 2026.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-18T07:35:18.103873+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/17/anthropics-annualized-revenue-surges-to-65b/"},{"assertion_indexes":[4],"citation_authority_claim_id":48000,"citation_checks":[],"claim_id":48000,"claim_text":"Anthropic's investors expect the company's revenue to finish 2026 between $100 billion and $120 billion, as reported by the Financial Times.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-18T07:35:18.103873+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/17/anthropics-annualized-revenue-surges-to-65b/"},{"assertion_indexes":[5,11],"citation_authority_claim_id":50930,"citation_checks":[],"claim_id":50930,"claim_text":"OpenAI is hiring forward-deployed engineers in San Francisco with a salary range of $162,000 to $280,000 plus equity.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:34:20.763273+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/become-forward-deployed-engineer"},{"assertion_indexes":[5,11],"citation_authority_claim_id":50931,"citation_checks":[],"claim_id":50931,"claim_text":"OpenAI currently has nineteen open requisitions for forward-deployed engineer positions.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:34:20.763273+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/become-forward-deployed-engineer"},{"assertion_indexes":[5,11],"citation_authority_claim_id":50932,"citation_checks":[],"claim_id":50932,"claim_text":"Handshake has posted a senior version of the forward-deployed engineer role with a salary range of $250,000 to $350,000.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:34:20.763273+00:00","source_name":"Nate Jones","source_url":"https://natesnewsletter.substack.com/p/become-forward-deployed-engineer"},{"assertion_indexes":[6],"citation_authority_claim_id":39855,"citation_checks":[],"claim_id":39855,"claim_text":"There are approximately 17,000 forward-deployed engineers currently on the U.S. market, with a significant portion employed by Palantir.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-31T06:25:22.075505+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/07/30/forward-deployed-engineers-are-the-ai-industrys-latest-talent-obsession/"},{"assertion_indexes":[7,13],"citation_authority_claim_id":50961,"citation_checks":[],"claim_id":50961,"claim_text":"The European Commission's AI Office reported that analyzing model usage can serve as part of post-market monitoring under Measure 3.5 of the GPAI Code of Practice.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:35:20.652368+00:00","source_name":"EU AI Act Newsletter","source_url":"https://artificialintelligenceact.substack.com/p/the-eu-ai-act-newsletter-109-the"},{"assertion_indexes":[7,13],"citation_authority_claim_id":50962,"citation_checks":[],"claim_id":50962,"claim_text":"The AI Office stated that 'marginal-risk' clauses in the GPAI Code of Practice allow providers to deploy unsafe models if competitors do so, but only in exceptional circumstances with appropriate safeguards.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:35:20.652368+00:00","source_name":"EU AI Act Newsletter","source_url":"https://artificialintelligenceact.substack.com/p/the-eu-ai-act-newsletter-109-the"},{"assertion_indexes":[7,13],"citation_authority_claim_id":50963,"citation_checks":[],"claim_id":50963,"claim_text":"The EU AI Act Service Desk clarified that AI agents are not a separate regulatory category under the Act but are covered by existing definitions of AI systems and general-purpose AI models.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:35:20.652368+00:00","source_name":"EU AI Act Newsletter","source_url":"https://artificialintelligenceact.substack.com/p/the-eu-ai-act-newsletter-109-the"},{"assertion_indexes":[8],"citation_authority_claim_id":22671,"citation_checks":[],"claim_id":22671,"claim_text":"The European Commission has released a plan to manage the risks and opportunities of advanced AI in cybersecurity, involving collaboration with member states, industry, and EU organizations to build evaluation capacity and a secure testing platform.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T07:11:31.122277+00:00","source_name":"EU AI Act Newsletter","source_url":"https://artificialintelligenceact.substack.com/p/the-eu-ai-act-newsletter-106-calls"},{"assertion_indexes":[8],"citation_authority_claim_id":22683,"citation_checks":[],"claim_id":22683,"claim_text":"Legal experts argue that the EU AI Act's existing definitions and requirements are likely to apply to agentic AI systems due to their autonomy and adaptiveness.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T07:11:31.122277+00:00","source_name":"EU AI Act Newsletter","source_url":"https://artificialintelligenceact.substack.com/p/the-eu-ai-act-newsletter-106-calls"},{"assertion_indexes":[8],"citation_authority_claim_id":22684,"citation_checks":[],"claim_id":22684,"claim_text":"Compliance with the EU AI Act is expected to be more difficult for agentic AI systems because their ability to change after deployment may trigger 'substantial modification' rules and tension with human oversight requirements.","correction_role":null,"label":"opinion","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-07-14T07:11:31.122277+00:00","source_name":"EU AI Act Newsletter","source_url":"https://artificialintelligenceact.substack.com/p/the-eu-ai-act-newsletter-106-calls"},{"assertion_indexes":[9,14],"citation_authority_claim_id":50920,"citation_checks":[],"claim_id":50920,"claim_text":"Judge William Alsup ordered Anthropic to pay a $1.5 billion copyright settlement to writers whose works were used to train its AI models.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:34:10.690022+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/23/is-it-legal-to-train-ai-models-on-copyrighted-books-its-complicated/"},{"assertion_indexes":[9,14],"citation_authority_claim_id":50921,"citation_checks":[],"claim_id":50921,"claim_text":"Judge William Alsup ruled that Anthropic's AI training was lawful, penalizing the company only for pirating books from illegal online shadow libraries.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:34:10.690022+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/23/is-it-legal-to-train-ai-models-on-copyrighted-books-its-complicated/"},{"assertion_indexes":[9,14],"citation_authority_claim_id":50922,"citation_checks":[],"claim_id":50922,"claim_text":"Judge Stephanos Bibas ruled last year that it was not fair use to train on Thomson Reuters' content to build a competing AI-based legal platform.","correction_role":null,"label":"fact","paired_claim_ids":[],"provenance_tier":"primary","recorded_at":"2026-08-24T06:34:10.690022+00:00","source_name":"TechCrunch AI","source_url":"https://techcrunch.com/2026/08/23/is-it-legal-to-train-ai-models-on-copyrighted-books-its-complicated/"}],"version":"evidence-bundle-v1"}