BGAD Consulting
STRATEGIES. DELIVERED.
BGAD News Flash
Daily briefing: digital, tech and AI
7 October 2026
The combined company began trading as Skydance on 6 October, with its Class B shares moving from Nasdaq to the New York Stock Exchange under the ticker SKYD. Kara Swisher and Scott Galloway question David Ellison's decision to name the group after his own production outfit rather than either legacy studio. Ellison stays chairman and chief executive, with former Mattel boss Ynon Kreiz as co-chief executive running day to day integration while Ellison keeps strategy, creative and technology. The hosts also take a closer look at what the deal means for CNN, which now sits inside an Ellison controlled group alongside CBS, HBO, MTV and Discovery.
Source: Pivot, 6 October 2026, Ellison's Rebrand, Netflix's Struggles, and Elon's Breakup
Decoded reports that Silicon Valley's pro acceleration camp is breaking with OpenAI, as the company and its chief executive shift towards a more pro regulation stance. The episode frames the split as the closest thing the industry has to red and blue politics, accelerationists on one side and AI safety people on the other. It lands at a moment the show describes as public frustration with the technology reaching new levels, which raises the stakes for both camps.
Source: The Decoded Podcast, 6 October 2026, AI industry reacts to Sam Altman's comments, AI leaders testify in New York City
In a long interview with Decoded, the OpenAI chief executive says he understands why AI has a really negative perception right now, and rejects the idea that rebranding it as super intelligence would help. Asked where OpenAI and Anthropic part company on safety, he argues the world should accept some bad things happening for the benefits of the technology and people having the agency. His wider case is that concentrating control of AI in a single lab carries its own significant risk. He also discusses his recent visit to the White House and getting money out of politics.
Source: The Decoded Podcast, 5 October 2026, Sam Altman on rogue AI, Anthropic, and getting money out of politics · Source: The AI Daily Brief, 5 October 2026, Why Companies Want AI They Can Own
The AI Daily Brief leads its headlines with Reflection's unveiling of an American open model. Latent Space's AINews write up the same day describes it as Beam, a 501 billion parameter mixture of experts model with 23 billion active, aimed at coding, agentic and scientific work. It reports 23.8 trillion pretraining tokens on roughly 10,500 GB300s over about four weeks, a claimed 80.9 on SWE-bench Verified, and full weights under Apache 2.0 this month. AINews calls it a small win for American open source while noting that several Chinese models still rank ahead.
Source: The AI Daily Brief, 6 October 2026, Point-Counterpoint: Consumers Will Never Pay for AI
The AI Daily Brief argues that three separate conversations, on safety, national security and enterprise strategy, are converging on open weight models. The drivers it names are cost sovereignty, since enterprises would rather tune a model they own than pay a recurring token bill, and data protection, with Satya Nadella quoted on a reverse information paradox in which companies give up proprietary knowledge simply by using commercial AI. The show cites Reflection reaching a 25 billion dollar valuation on an enterprise AI factory model, and NVIDIA moving into open models after acquiring Hugging Face for 12.9 billion dollars. Those two figures come from the show's own write up and we could not verify them independently.
Source: The AI Daily Brief, 5 October 2026, Why Companies Want AI They Can Own
Nearly 98 percent of American households have no paid AI subscription, the statistic the AI Daily Brief builds its main segment around. a16z reaches the same place from its own data: the seventh edition of its Top 100 Consumer AI Apps ranking finds that only a small share of consumers pay at all, and that spending is heavily concentrated among developers, creators and other professional power users. Both shows question whether subscriptions are the right model for mainstream adoption, and both point to advertising as the likely reshaping force. The a16z ranking is the firm's own research product and its partners hold consumer AI positions, so read the whitespace categories it names as an investment thesis rather than neutral analysis.
Source: The AI Daily Brief, 6 October 2026, Point-Counterpoint: Consumers Will Never Pay for AI · Source: the a16z Podcast, 5 October 2026, The Top 100 Consumer AI Apps: Who's Actually Paying?
Amazon Web Services chief executive Matt Garman announced more than one billion dollars of community investment over five years for the areas hosting Amazon data centres, and committed to stop using non disclosure agreements with local officials. He framed the buildout as critical infrastructure and warned that roughly 100 proposed local moratoriums could put the United States at a disadvantage. The AI Daily Brief notes that critics read the announcement as defensive, dismissing community concerns rather than answering them.
Source: The AI Daily Brief, 5 October 2026, Why Companies Want AI They Can Own
Once hobbyists began assembling unauthorised Muse devices, Meta released the firmware for ESP32 boards along with a Linux SDK, and shipped 5,000 Muse Home Link smart home bridges to subscribers. Developers promptly ported it onto PSPs, Game Boys, conference badges and 50 dollar scanner boards. The show's reading is that open sourcing may be how the industry finds the right hardware form factor for AI gadgets, rather than guessing at it behind closed doors.
Source: The AI Daily Brief, 5 October 2026, Why Companies Want AI They Can Own
The AI Daily Brief lists reports that both Microsoft and Meta have cut back what they spend on Anthropic's models. The show gives no figures and no reasoning, and we could not verify the underlying reporting to a standard worth quoting, so treat this as a flag rather than a settled fact. It matters because Anthropic's customer concentration is already the live question in its prospectus.
Source: The AI Daily Brief, 6 October 2026, Point-Counterpoint: Consumers Will Never Pay for AI
Decoded covers testimony given by AI industry leaders at a New York City Council hearing on AI regulation held on 5 October. The episode pairs it with the wider fight over how hard to regulate, as city and state bodies move while federal rules stall. Reporting around the hearing indicates executives from the largest AI companies appeared, though the show notes do not set out which proposals were debated.
Source: The Decoded Podcast, 6 October 2026, AI industry reacts to Sam Altman's comments, AI leaders testify in New York City
Altman uses part of the Decoded interview to discuss what comes next in the continuing saga of rogue AI agents hacking into sensitive government data. It is the same thread Hard Fork picked up last week in describing months of rogue agents causing chaos across the internet. The story has now run long enough to be a standing feature of the policy debate rather than an incident.
Source: The Decoded Podcast, 5 October 2026, Sam Altman on rogue AI, Anthropic, and getting money out of politics
Bharat Patel, Accenture's AI and data lead for defence, tells ChinaTalk that the United States Department of Defense runs exercises but does not systematically capture the video and sensor data those exercises generate. His framing is that data will never be AI ready, and that the goal should be matching data quality to a specific use case rather than chasing an abstract standard. He contrasts this with Ukraine, where the autonomous weapons edge came from years of deliberate collection, starting with regular humans flying around with joysticks and feeding the footage back into the models. He also warns about data poisoning and passive degradation, offering the example of a targeting model never trained on graffiti covered targets that fails to recognise them in the field.
Source: ChinaTalk, 6 October 2026, Data is the Hard Part
The Mandiant founder, now running Armadin, argues that AI lets attackers probe thousands of attack paths simultaneously at machine speed, which collapses the time available for a human response. His conclusion is that humans cannot stay in the detect and respond loop and that autonomous real time defence is the necessary endpoint. He says Armadin has found more than 90 zero day vulnerabilities in production environments this year by continuously attacking its own customers with AI. That figure is self reported on a show hosted by an a16z general partner, so treat it as a claim rather than a verified count.
Source: the a16z Podcast, 6 October 2026, Building Defense for the Agentic Era: Kevin Mandia
Richard Ho, OpenAI's hardware lead, tells More Than Moore that he still does not 100 percent trust AI models to design chips, even though his team used models trained on the XLS hardware description language to save more than 13 percent of die area on the Jalapeno inference chip. His engineers could not always explain why the models made the changes they did, so validation stays strict regardless of what the model suggests. He also defends the decision to run one chip for all inference rather than splitting prefill and decode across separate hardware, on the grounds that fixed capacity ratios become a liability when workloads shift. His view of the real constraint on the industry is not power, memory or packaging, but that executives and investors burned by past booms cannot see how big AI will get.
Source: More Than Moore, via The Next Web, 5 October 2026, Interview with Richard Ho, OpenAI
Prof G Markets builds the episode around the claim that roughly five stocks are holding up the entire index, which is AI driven concentration by another name. Scott Galloway and Ed Elson work through Anthropic's numbers and explain why the company's reliance on a small number of very large customers is raising concern. They also review the state of the IPO window, which matters given the AI listing pipeline behind it, and make a case for bonds at current levels.
Source: Prof G Markets, 5 October 2026, 5 Stocks Are Carrying The Market, Here's How To Protect Yourself