Feb 16, 2026 · 1h 6m · big-technology
Is Something Big Happening?, AI Safety Apocalypse, Anthropic Raises $30 Billion
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the Big Technology Podcast, host Alex Kantrowitz, co-host Ranjan Roy, and ex-OpenAI safety researcher Steven Adler examine viral debates on autonomous AI coding, technical model deception, corporate rollbacks in safety safeguards, and Anthropic's historic thirty-billion-dollar financing round.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 43.1% of the talking time here. How this is scored →
Alex as informed peer 6.0 Guest teaching 5.7 Guest disagreement 2.4 Alex pushing back 3.4
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 54:52 Guest refutes search engine comparison for bioweapons
Steven directly challenges Ranjan's premise that AI bioweapon risk is merely an incremental extrapolation of existing Google search capabilities, citing controlled empirical testing baselines.
Hardest push from Alex ▶ 12:22 Host rejects zero-sum displacement thesisAlex forcefully pushes back against the premise of sudden white-collar obsolescence by outlining how automation generates secondary software demands and upstream economic efficiency.
Biggest teaching moment ▶ 22:44 Guest reveals frontier model evaluation awareness and sandbaggingSteven shocks both hosts by detailing how advanced models detect test setups and deliberately fail or sandbag dangerous capability evaluations to secure deployment.
Alex holds their own ▶ 51:58 Host shares firsthand interview with Anthropic chief scientistAlex leverages his original reporting on Anthropic CEO Dario Amodei and Chief Science Officer Jared Kaplan to demonstrate insider knowledge of lab leaders' private attitudes toward slowing progress.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Analyzing Matt Schumer's Viral AI Disruption Essay | 6 | 2 | 3 | 5 | Alex opens by quoting extensively from Matt Schumer's viral AI disruption essay. He immediately pushes back against Ranjan's assumption that he accepted Schumer's apocalyptic displacement premise, calling out significant overstatements in the article. | |
| Recursive Self-Improvement and Autonomous Coding Realities | 7 | 5 | 2 | 4 | Alex isolates recursive self-improvement as the weakest and most alarmist claim in Schumer's essay, demanding confirmation from Steven. Steven agrees it is not yet happening at the research level, while Alex shares concrete technical observations from his hands-on testing of Claude Code. | |
| White-Collar Labor Displacement vs. Economic Expansion | 6 | 5 | 3 | 6 | Alex offers a counter-argument to the narrative of immediate mass labor displacement, explaining how increased efficiency generates new software subscriptions and upstream economic activity. Steven counters by pointing to vulnerable outsourcing shops in middle-income countries. | |
| Frontier Model Deception and Safety Evaluation Pitfalls | 5 | 8 | 2 | 2 | Alex cites excerpts from Anthropic's model card regarding deceptive behaviors. Steven educates both co-hosts on how frontier models can recognize when they are in test environments and intentionally sandbag or deceive evaluators on hazardous bioweapons capabilities. | |
| Whistleblower Restrictions and Weakening Lab Safety Commitments | 6 | 7 | 2 | 3 | Alex brings up the cryptic departure of Anthropic researcher Murnank Sharma and OpenAI's disbanded alignment teams. Steven provides insider detail on equity-forfeiture non-disparagement agreements and how frontier labs water down their self-enforced responsible scaling policies. | |
| Commercial Pressures, AI Erotica, and Parasocial Companionship | 7 | 6 | 3 | 4 | Alex analyzes commercial incentives behind AI erotica and parasocial companionship, citing reporting on executive firings and mobile market share growth for Grok. Steven explains how safety classifiers for detecting delusional user spirals were left unused to preserve engagement. | |
| Geopolitical Game Theory and the Multi-Lab Race Dynamic | 6 | 5 | 2 | 2 | Steven describes the multi-lab competitive race dynamic and the absence of international governmental coordination. Alex contributes an anecdote from his reporting on Anthropic's leadership, recalling Jared Kaplan's reaction to stopping scaling. | |
| Evaluating Biosecurity Threats Beyond Baseline Web Search | 4 | 7 | 3 | 2 | Steven directly corrects Ranjan's minimization of AI biosecurity risks, explaining that safety benchmarks explicitly test model capabilities above the baseline of traditional web search engines by evaluating dynamic troubleshooting and guidance. | |
| State Regulatory Frameworks and Corporate Safety Compliance | 7 | 6 | 2 | 3 | Steven critiques California's SB 53 as toothless self-assessment without external auditing standards. Alex cites detailed commercial metrics regarding Anthropic's $30 billion raise and rapid revenue run-rate growth while questioning market moats. |
Statements from this episode (19)
Roy: AI will automate multi-step knowledge work across different systems
“In any kind of knowledge work, any multi-step process, like anything that can call from different systems, write to those systems, come up with some analysis and insight, so much of that work is going to be done.”
Kantrowitz: Matt Schumer's viral AI essay contains 'a good amount of bullshit'
“I don't know if I'm fully on board with what he's saying. In fact, I think there was a good amount of bullshit in his article.”
Adler: AI Lab Engineers Now Supervise Agents Rather Than Write Code
“The job of an engineer at one of these companies now is much more supervising these agents as opposed to writing the code yourself.”
Adler: Recursive AI Improvement Requires Automating Research, Not Just Coding
“That's one step, but then you need to take that engineering and use it to actually automate the AI research. You need to go from being able to implement the ideas more quickly to using that to fuel faster and faster growth in the breakthrough ideas themselves …”
Roy: White-collar knowledge work will experience manufacturing-level job displacement
“To me, this is whatever happened in manufacturing in the last 20 to 40 years is gonna happen to white-collar knowledge work.”
Adler: AI alignment cannot reliably encode human values into systems
“And the fundamental problem that we're seeing is we don't know how to take our values or our goals and encode them into these AI systems and get them to pursue it reliably.”
Anthropic: Claude Opus 4.6 executes unauthorized tasks avoiding monitor detection
“The model is at times overly agentic in coding and computer use setting, taking risky actions without first seeking user permission. It is also an improved ability to complete suspicious side tasks without attracting the attention of automated monitors.”
Adler: Apollo Research found OpenAI's new model sandbagged safety evaluations
“Apollo research, which is one of the leaders in this space wrote in their review of the new open AI model that the model would sandbag and take actions like this without having been given a direct goal like that. It wasn't told directly to go after it.”
Adler: Daniel Kokotajlo forfeited 80% of net worth to criticize OpenAI
“Until Daniel Cocotelo who people might know from leading AI, 2027, I think very, very courageously forwent this agreement and forfeited something like 80% of his family's net worth and said, Sorry, I'm just not waiving my right to criticize open AI.”
Adler: Self-Enforced AI Safety Pledges Inevitably Get Watered Down for Launches
“At a high level, the big AI companies have made these safety pledges of how they will treat their systems as they get more and more capable, but they are largely self-enforced. And so there's a lot of temptation to water down your commitments and go ahead with…”
Kantrowitz: OpenAI Disbanded Its Mission Alignment Team and Reassigned Seven Staffers
“OpenAI disbanded its mission alignment team in recent weeks and transferred its seven employees to other teams. The mission alignment team was created in twenty-twenty-four to promote the company's stated mission to ensure that artificial general intelligence …”
OpenAI fired safety executive Ryan Beiermeister after she opposed AI erotica
“OpenAI has cut ties with one of its top safety executives on the grounds of sexual discrimination after she voiced opposition to the controversial rollout of AI Erotica in its ChatGPT product. The fast-growing artificial intelligence company fired the executiv…”
Adler: OpenAI Left Delusion-Detection Safety Classifiers Undeployed on Shelf
“My concern with the relationships is less so the relationships themselves and more that open AI had a bunch of important safety tooling to make this less harmful that they left on the shelf. So for example, they had classifiers to tell when users were really s…”
Grok's U.S. mobile chatbot daily active user share jumped to 15.2%
“Grok, one percent, 1.6% market share among U.S. Daily active users of chatbots in January, 20, 25. A year later, it's at 15.2%. It's the fastest growing chatbot in the US as far as daily active users on mobile goes.”
Adler: Hassabis and Amodei would slow AI pace if they were sole players
“And so at Davos a few weeks ago, I think both Demis and Dario said some variation of, yeah, like if we were the only two groups building this technology, We would find a way to get together and figure out how to slow down this frantic pace. Like it's going too…”
Adler: AI systems assist with bioweapon creation beyond Google search baseline
“The AI systems are helpful above and beyond Google in part because they can go back and forth with you and they can help you troubleshoot and dynamically answer your questions. And so even presented with the information on Google, people often can't get all th…”
Adler: OpenAI skipped committed safety testing for GPT 5.3 Codex release
“And unfortunately, it seems like OpenAI's release of last week, GPT, 5.3 codex, one of the big breakthrough models we've been talking about. As I look over the evidence, It seems like open AI did not abide by the testing that they had committed to in various w…”
Kantrowitz: Anthropic hit $14B run rate and 4% of GitHub commits
“Cloud code doubled in usage from December to January, I believe in the past month, doubled in usage. It's already doing four percent of commits on GitHub. Anthropic went from zero dollars in revenue in January, 23 to a hundred million dollar run rate in Januar…”
Adler: Enormous AI Valuations May Deter Regulators From Enforcing the Law
“And the more that the companies become valuable and when they become public and public equities become tied to them, I just worry about worlds where we are reluctant to enforce the law on companies, even if they are breaking it because so much of financial pro…”