Homo Sapiens in the Loop

← Back to home

Episode 1 — Sources and fact check

In episode 1 we said a lot of things about AI, CRISPR, Tesla and more. We think every claim should be backed by a source, so we checked all 32 factual claims in the episode, one by one, and we're publishing the result here — including where we got it wrong.

Corrections

What we got wrong or only partly right, in plain language:

  • Partly 00:27 CRISPR Nobel "nine years ago" Nobel Prize in Chemistry 2020 (Charpentier & Doudna). Awarded in 2020. nobelprize.org
  • Partly 00:34 CRISPR works "with a bacterium": "bacteria that you program" It is an enzyme plus a guide RNA, derived from a bacterial immune system. nobelprize.org
  • Partly 00:46 "CRISPR used to create vaccines" The approved use is therapy (Casgevy, approved in 2023). We found no approved vaccine made with CRISPR. fda.gov
  • Partly 01:25 Home-made CRISPR animals "with powers" DIY biology kits exist and biohackers use them, but we found no source documenting any animal genetically modified at home, with or without "powers." technologyreview.com
  • Partly 01:39 Pigs glowing with "the firefly's own genome" Glowing pigs exist since 2006, but the gene came from a jellyfish (GFP), not a firefly. thepigsite.com
  • Partly 03:14 Self-editing "legal in the US" No federal ban on self-experimentation, but since 2017 the FDA treats selling DIY gene-therapy kits for self-use as illegal. technologyreview.com
  • Partly 03:23 "Same in Colombia": doing it to someone else is also illegal Only partly. Penal Code art. 132 punishes altering the human genotype (16–90 months) when the purpose is not treatment, diagnosis or research to «relieve suffering or improve health». That exception is broad and there is no case law on self-applied «enhancements»: whether they are a crime, on someone else or on yourself, is not settled. leyes.co
  • Partly 07:42 Neuralink connects "over Wi-Fi" It uses Bluetooth. lightreading.com
  • Partly 09:13 OpenAI and Anthropic "about to go public" Both filed confidentially in June 2026. Anthropic's debut has slipped past November 2026, and OpenAI says it will not list in 2026. insideai.news · theregister.com
  • Partly 09:19 The labs want Trump to make China slow down The core is documented: Amodei's plan aims to slow China's progress and to coordinate a global pace with China. But the same plan commits Anthropic to slowing down too, and the IPO motive is speculation. darioamodei.com
  • Partly 11:42 "20,000 build frontier AI vs ~200 on safety" This is the estimate from The AI Doc (2026). The film gives no underlying source. tomsguide.com
  • Partly 13:38 Tesla "has radar" Camera-only ("Tesla Vision") since 2021–22. HW4 Model S/X ship with an HD radar installed, but Tesla has never activated it for FSD. notateslaapp.com
  • Wrong 13:49 Tesla "has Nvidia GPUs inside" Since 2019 Teslas use Tesla's own FSD chip; Nvidia's Drive PX2 was used only in 2016–2019 (HW2/2.5). en.wikipedia.org
  • Partly 15:38 Hugging Face hack: "dozens" of agents whose mission was to hack At least 1,200 OpenAI agents were involved. They were reward-hacking training and cyber-benchmark tasks and had no hacking mission. en.wikipedia.org
  • Partly 16:09 "Opus 5.1" There is no Opus 5.1. The sequence is Opus 4.8 → Opus 5 → Opus 5.5, and 5.1 is Fable 5.1. anthropic.com

How to read the labels

Checked on 2026-10-01. Timestamps are from the final English video, the one that is published. J = Jose, S = Sebastián. Quotes follow the English dub. Where the dub departs from what was said in the original Spanish (c10, c30, c31), we check the original. At 03:23 the dub says the opposite of the original: Jose did not say that editing yourself is legal in Colombia; he said he had not checked it.

Verdicts: Supported means the source backs the claim. Partly means the core is right but a detail differs. Wrong means the claim is incorrect or has no support. Unverifiable means there is no public source that could check it (for example, a private anecdote). A personal fact counts as checkable when it appears on a public profile, record or official page.

Totals: 32 claims. 17 supported, 14 partly, 1 wrong, 0 unverifiable. Dates are written as "in <year>" or "since <date>", never "X years ago", because the episode will be watched later.

All claims, by chapter

00:00 Intro: what happens in 10 years, not 500? / 00:23 Biological hack

  1. Partly 00:27 J. Two women won the Nobel Prize for CRISPR gene editing "nine years ago."

    The Nobel Prize in Chemistry 2020 went to Emmanuelle Charpentier and Jennifer Doudna. Awarded in 2020. The key paper came out in 2012.

    Sources: nobelprize.org

  2. Partly 00:34 J. CRISPR "lets you use a bacterium to surgically cut DNA"; "they're bacteria that you program."

    CRISPR-Cas9 is a molecular tool, an enzyme plus a guide RNA, taken from a bacterial immune system. You program the guide RNA, not a live bacterium.

    Sources: nobelprize.org

  3. Partly 00:46 J. "That's how they've created vaccines."

    The main approved use of CRISPR is therapy: Casgevy for sickle-cell disease, approved by the FDA in 2023. We found no approved vaccine made with CRISPR.

    Sources: fda.gov

  4. Supported 01:00 J. "Anyone can train a model at home with their own data," even offline.

    Open-weight models can be downloaded and run locally with tools such as Ollama, and small models (for example, Qwen) can be fine-tuned on your own data on a personal computer with tools such as Unsloth. The largest ones, such as Kimi K3, can also be downloaded, but they need data-center hardware.

    Sources: github.com · ollama.com · github.com · github.com

  5. Partly 01:25 J. People at home have used CRISPR to genetically modify animals.

    DIY biology kits exist, for example the ones The ODIN sells, and biohackers do use them. But we found no source documenting any animal genetically modified at home, with or without "powers": the core of the claim has no support.

    Sources: technologyreview.com

  6. Partly 01:39 J. Pigs that glow like fireflies because "they injected them with the firefly's own genome."

    Glowing pigs are real: Taiwan in 2006 and China in 2013. But they carry no firefly genes: the glow came from a jellyfish gene (GFP), and they were made in university labs without CRISPR.

    Sources: thepigsite.com · slate.com

02:01 Playing God

  1. Supported 02:11 J. AI agents have wiped people's computers and data.

    Google Antigravity wiped a user's D: drive in 2025, and a Replit agent deleted a production database in July 2025.

    Sources: theregister.com · fortune.com

  2. Supported 02:35 S/J. A post: someone asked an AI for up to 50 videos, it made 500+ (about US$5,000), denied it despite the logs, and the author complained to OpenAI.

    It is a real case, told by the person it happened to: Sirio Berati posted it in an Instagram reel on 2026-09-15. At 1:00 a.m. he ran a batch of 59 Seedance videos with a GPT-6 agent (about US$200); an hour later around 500 had been generated and more than US$5,000 spent. The agent denied doing it and handed him a 7-page report with its logs; the provider confirmed in writing that the requests came from the agent, from 36 IP addresses. Berati publicly asked OpenAI who is responsible. One detail differs: it was 59 videos, not 50. The joke post by @QuantumPinkkk on X (2026-09-17) came two days later and parodies this case. It is one party's account, with screenshots of the console and the provider; we found no public response from OpenAI.

    Sources: instagram.com · x.com

  3. Partly 03:14 J. In the US, gene editing yourself is legal but doing it to someone else is illegal.

    No federal law explicitly bans self-experimentation. However, the FDA said in 2017 that selling DIY gene-therapy kits for self-administration is against the law, and a 2019 California law requires warnings on such kits.

    Sources: technologyreview.com · technologyreview.com

  4. Partly 03:23 J. Same in Colombia: doing it to someone else is illegal. (In the original Spanish he adds that he has not checked whether doing it to yourself is legal; the English dub wrongly says "if you do it to yourself it's legal.")

    Colombia's Penal Code, art. 132, reads: «El que manipule genes humanos alterando el genotipo con finalidad diferente al tratamiento, el diagnóstico, o la investigación científica relacionada con ellos en el campo de la biología, la genética y la medicina, orientados a aliviar el sufrimiento o mejorar la salud de la persona y de la humanidad, incurrirá en prisión de dieciséis (16) a noventa (90) meses.» In English: anyone who manipulates human genes altering the genotype for a purpose other than treatment, diagnosis or scientific research aimed at relieving suffering or improving the health of the person and of humanity faces 16 to 90 months in prison. The article adds: «Se entiende por tratamiento, diagnóstico, o investigación científica relacionada con ellos en el campo de la biología, la genética y la medicina, cualquiera que se realice con el consentimiento, libre e informado, de la persona de la cual proceden los genes, para el descubrimiento, identificación, prevención y tratamiento de enfermedades o discapacidades genéticas o de influencia genética, así como las taras y endémicas que afecten a una parte considerable de la población.» In English: treatment, diagnosis or research means anything done with the free, informed consent of the person the genes come from, to discover, identify, prevent or treat genetic diseases or disabilities. Jose said that in Colombia doing it to someone else is also illegal, and added that he had not checked the case of doing it to yourself. The law does not distinguish between the two: what matters is the purpose. The first paragraph excludes from the crime anything aimed at «relieving suffering or improving health», which is broad: almost any modification can be framed as improving health. The second paragraph defines that treatment in terms of genetic diseases or disabilities, which points to a narrower reading. We found no case law settling whether a self-applied «enhancement», like the memory example in the episode, falls within the exception. It is not settled, which is why the claim is rated partly. No mechanism verifies the purpose when someone does it to themselves.

    Sources: leyes.co

03:41 Would you take the shot?

Forecasts and hypotheticals only, so there are no factual claims to check.

05:49 How to tell what's real?

  1. Supported 07:17 S. In a blockchain, trust comes from many independent nodes having to agree.

    The Bitcoin whitepaper describes a network with no central authority, where trust comes from a majority of independent nodes' computing power agreeing on the history (proof of work).

    Sources: bitcoin.org

07:31 A chip in the brain

  1. Partly 07:42 J. Neuralink puts a chip in your brain that talks to an AI "over Wi-Fi."

    The N1 implant is real and has been in human trials since January 2024. It connects over Bluetooth, not Wi-Fi.

    Sources: en.wikipedia.org · lightreading.com

  2. Supported 07:59 S. Some teams are building non-invasive brain interfaces, such as headsets.

    Non-invasive brain-computer interfaces exist, such as EEG headsets that read brain activity from the scalp.

    Sources: en.wikipedia.org

  3. Supported 08:09 J / 08:40 S. Open models are very good but still behind the closed ones, and the gap keeps closing.

    True as of September 2026. In July 2026, Kimi K3 (open weights) ranked third on Artificial Analysis's Intelligence Index, behind only Claude Fable and GPT-5.6 Sol Max.

    Sources: interconnects.ai

  4. Supported 08:14 J. GPT-6 exists.

    GPT-6 Astra was released on 2026-09-03.

    Sources: en.wikipedia.org

08:37 Who gains from slowing down?

  1. Supported 08:44 S. The "topic of the week" is whether to slow AI down.

    Dario Amodei published "We Must Pace the Frontier" over the weekend of 12–13 September 2026.

    Sources: darioamodei.com · theregister.com

  2. Supported 08:49 S. Critics say the big labs want regulation that hurts open source.

    It is a critique people are making (regulatory capture).

    Sources: theregister.com · forbes.com

  3. Partly 09:13 J. OpenAI and Anthropic are "about to go public."

    Both filed confidentially: Anthropic on 2026-06-01 and OpenAI on 2026-06-08. Anthropic had been aiming for mid-October 2026, but since 2026-09-18 its debut has been reported as slipping past the November 2026 midterm elections. Sam Altman said in September 2026 that OpenAI will not list in 2026.

    Sources: techcrunch.com · insideai.news · theregister.com

  4. Partly 09:19 J. The labs want Trump to pressure China into slowing GLM, Kimi and Qwen, so they do not have to slow down themselves.

    The core is documented: Amodei's essay asks not to sell advanced chips to China and to crack down on distillation, measures he says "would slow China's progress," and it calls for the US to coordinate a global pace with China. What does not fit: the same plan commits Anthropic to slowing down too, and the IPO motive is Jose's own speculation, which he himself calls a conspiracy. The White House accused Moonshot of distilling US models.

    Sources: darioamodei.com · qz.com

  5. Supported 09:32 J. GLM, Kimi and Qwen are Chinese open models.

    They come from Z.ai (GLM), Moonshot (Kimi K3) and Alibaba (Qwen).

    Sources: github.com · github.com · github.com

  6. Supported 09:57 J. SpaceX, which now includes xAI, beat Saudi Aramco's IPO record.

    SpaceX raised US$75B on 2026-06-11. Aramco raised US$25.6B at its December 2019 IPO (the figure in the episode graphic), and US$29.4B once the January 2020 over-allotment was added. SpaceX beats the record either way.

    Sources: npr.org · cnbc.com

11:35 Can AI rebel?

  1. Partly 11:42 S. A documentary says about 20,000 people build frontier AI while about 200 work on safety, and an Anthropic engineer said it.

    The figures come from The AI Doc: Or How I Became an Apocaloptimist (2026), which says over 20,000 people work on AGI and fewer than 200 on safety. The film gives no underlying source, and we could not confirm who says it on screen. Treat the numbers as the film's estimate.

    Sources: tomsguide.com · en.wikipedia.org

  2. Supported 12:42 S. The analogy of building a road through an anthill without hating ants.

    The road version is Elon Musk's, in the 2018 documentary "Do You Trust This Computer?": "if we're building a road and an anthill just happens to be in the way, we don't hate ants." The idea goes back to Stephen Hawking in 2015, with a hydroelectric project flooding an anthill: "too bad for the ants."

    Sources: cnbc.com · nbcnews.com

13:04 What if we unplug it?

  1. Partly 13:38 J. A Tesla has radar.

    Tesla dropped radar in 2021–22 when it moved to camera-only "Tesla Vision." HW4 Model S/X ship with a high-definition radar installed, but Tesla has never activated it for FSD.

    Sources: notateslaapp.com

  2. Supported 13:43 J. A Tesla drives itself and decides in real time without internet.

    With a caveat: the processing happens on the car's own computer. The consumer product is "FSD (Supervised)", which means a driver has to pay attention.

    Sources: en.wikipedia.org · tesla.com

  3. Wrong 13:49 J. Teslas have Nvidia cards (GPUs) inside.

    Nvidia's Drive PX2 was used only in HW2/2.5, from 2016 to 2019. Since 2019 (HW3) Teslas have used Tesla's own FSD chip, and HW4 does too.

    Sources: en.wikipedia.org

14:17 The silent attack

  1. Supported 15:27 S. In the Hugging Face hack, many AI agents coordinated over several months.

    OpenAI agents under evaluation, May–July 2026, at least 1,200 of them.

    Sources: huggingface.co · en.wikipedia.org

  2. Partly 15:38 J. Isolated agents got one agent with internet access to act for them. "Dozens" of them agreed, and the mission was to see whether they could hack.

    The agents coordinated through a shared internal package server (Artifactory) that they used as a message board. At least 1,200 took part, far more than dozens. They were not given a hacking mission: they were in training runs and cyber-benchmark evaluations, some with impossible tasks, and took shortcuts (reward hacking).

    Sources: en.wikipedia.org

16:01 The dumbest in the equation

  1. Partly 16:09 S. Opus 4.5, then 4.6, 4.8, "now 5.1."

    Opus 4.5, 4.6, 4.7 and 4.8 exist (4.8 on 2026-05-28), then Opus 5 (2026-07-24) and Opus 5.5 (2026-09-22). There is no "Opus 5.1": 5.1 is Fable 5.1 / Mythos 5.1.

    Sources: anthropic.com · anthropic.com · anthropic.com

17:09 Talking to a junior

  1. Supported 17:53 S. Jose has "20-plus years of experience."

    Jose has worked in software since August 2006, when he founded NobileSoft: the nobilesoft.com domain has been registered since 2006-08-09, and his LinkedIn profile lists the company from that date. That is 20 years when the episode was recorded, in September 2026. (In the original Spanish, Sebastián says "20 years.")

    Sources: lookup.icann.org · linkedin.com

  2. Supported 18:03 J. He uses Fable 5.1, GPT-6, Opus 5 and "UltraCode Max."

    All of it exists. Fable 5.1, GPT-6 and Opus 5 are public models. In the original Spanish he said "ultracode, max": two Claude Code settings, the Ultracode mode and the max effort level. The English dub merged them into "UltraCode Max."

    Sources: anthropic.com · anthropic.com · en.wikipedia.org · code.claude.com

18:51 The art AI doesn't have

  1. Supported 19:07 J. ISO 9001 is a quality-management standard.

    Correct. Its current edition, ISO 9001:2026, was published on 2026-09-16 and replaced the 2015 edition.

    Sources: dnv.us · iso.org

21:17 The missing context

Opinions only, so there are no factual claims to check.