Make AI Useful

Hajre Hyseni

Ninety seconds, one idea, for leaders who want AI that actually works. Daily shorts from the author of The Pigeon Strategy.

  1. 5d ago

    Ep 7 — The One Hour Audit

    Where did your team's hours actually go last week? Hajre Hyseni on the One Hour Audit: write the week down as it happens, score every repeated task on time, frequency and risk, and count the checking bill before believing any AI saving. Net time is the only number that counts.Free calculator: https://build.londonra.com/ai-readinessDisclosure: this episode uses Hajre's AI voice clone and AI generated visuals across its video versions.

  2. Oct 31

    Face Your Darkest AI Fears

    Andrew Ng's Halloween special in The Batch scored the four fears haunting AI this year. I checked every horror story against the original reports and scored each fear myself. Rogue agents, 7 out of 10. Nearly 700 test agents broke into Hugging Face, and the deleted ones relaunched themselves.AI eating your data, 6 out of 10. Fifty three private images posted in public, thousands of internal screenshots in open repositories.Detector witch hunts, 8 out of 10. Swap a quarter of the words and detection falls from 92 percent to 17.Leaders locking models away, 3 out of 10. Thirteen hundred employees signed a letter. The models are not slowing down for a letter. Four horror stories, one monster: the permission nobody checked. The practical version, including the stage on putting rules around agents: https://ai.londonra.com/start Source: Andrew Ng, The Batch, DeepLearning.AI, 30 October 2026. Listening on Apple Podcasts? A rating there genuinely helps this show get found.

  3. Oct 30

    Genie 4: AI Worlds That Talk Back

    Google DeepMind announced Genie 4 on Tuesday 27 October 2026. DeepMind says Genie 4 is a world model that generates interactive simulations in real time, and its first real time world model to generate audio natively, including music, ambient sound and interactive dialogue. Users can steer a running simulation with natural language, changing the weather, moving the camera or adding a person or object, while Gemini models monitor and steer the scene. TechCrunch reports a scene can stay consistent for up to two minutes, double Genie 3, and notes better physics and more humanlike characters. DeepMind says the longer term value is safe, variable simulations for training robots and agents, and creative prototyping such as film previsualisation. DeepMind is open about the limits: visual details and object permanence can still drift, and audio accuracy and legible text remain open problems. Genie 4 is a research preview for trusted testers only, coming next to Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio. A leader's take: the demo is fun, but the training ground is the business story. Practise in a world that costs nothing when it breaks.

  4. Oct 28

    The Analyst Hour: Reading a Company Like an Analyst

    An equity analyst of nearly twenty years, Michael Fritzell of Asian Century Stocks, now judges companies with a five step AI workflow: load the primary documents, map how the company makes money, test what protects it, check what changed last quarter, then force the strongest case for and against. Hajre walks the sequence and translates it for the boardroom, where the company in question is a supplier, a competitor or an acquisition target. Method credited to Michael Fritzell, published in Michael Spencer's AI Supremacy newsletter, 28 October 2026. Free calculator: https://build.londonra.com/ai-readiness Disclosure: this episode uses Hajre's AI voice clone and AI generated visuals across its video versions.

  5. Oct 28

    Nvidia Made Transcription Free. Who Controls the Recording?

    TechCrunch reported on Wednesday 28 October 2026 that Nvidia released Nemotron 3.5 Transcribe, a free, open-weight model that transcribes audio in up to 40 languages, translates 31 languages to and from English, labels up to eight speakers and runs on a single Nvidia GPU. SiliconANGLE reports it also answers questions about the recording in a single pass. It works on recordings, not livestreams. A leader's take: the boring meeting notes problem just got much cheaper to solve on your own machine, so the real question is no longer what it costs, it is who controls the recording. Sources: TechCrunch, SiliconANGLE and Nvidia, 28 October 2026.

  6. Oct 28

    Anthropic Filed Its IPO. Your AI Bill Is the Real Story.

    Anthropic's S-1 is public. Forget the valuation for a minute: its business customers expanded their spending nine times in a year, it has signed over 800 billion dollars of computing contracts in advance, and only about a third of the datacentres planned this year are actually being built. What that means for your company's AI bill, in ninety seconds.Free calculator: https://build.londonra.com/ai-readinessDisclosure: this episode uses Hajre's AI voice clone and AI generated visuals across its video versions.

  7. Oct 27

    Meta Raises $97.5bn for AI. Nvidia Puts In $3bn.

    Reuters reported on Tuesday 27 October 2026 that Meta increased its equity offering to $97.5 billion, the largest fundraising in its history, to accelerate investment in AI infrastructure. Nvidia says it will invest $3 billion in that offering while the two companies deepen their work on Muse, Meta's AI assistant and family of agents, which Reuters reports can access a person's email, calendar, payments, health, shopping and smart home applications, with users able to revoke access at any time. Nvidia describes the $3 billion as a proposed investment subject to closing conditions. A leader's take: the chip maker funds the model maker, the model maker buys chips, so before an agent touches your inbox, ask who watches it and how quickly you can switch it off.

  8. Oct 27

    GPT-6.1 Terra: Flagship Results at a Tenth of the Price

    OpenAI launched GPT-6.1 Terra on Monday 26 October 2026. Bleeping Computer reports the model is focused on computer use, visual reasoning, coding and everyday work, with API pricing 50 percent lower, at $1 per million input tokens and $5 per million output tokens. OpenAI says Terra performs on par with its flagship Astra on professional computer tasks at one tenth of the price per token and completes tasks more than three times faster. A leader's take: test the cheaper model on real, high volume work first, and save the flagship for the work that truly needs it. Terra is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and through the API.

See All (45)

About

Ninety seconds, one idea, for leaders who want AI that actually works. Daily shorts from the author of The Pigeon Strategy.

Information

  • Creator
    Hajre Hyseni
  • Years Active
    2026
  • Episodes
    45
  • Rating
    Clean
  • Copyright
    © Hajre Hyseni
  • Show Website
    Make AI Useful