lshaan Gupta

Building Agentic Workplace Wellness Companion | 3x Founder | NTU

[lshaan Gupta @ishaan80]

@OpenAI Completing the picture sooner than thought.

[lshaan Gupta @ishaan80]

@sama Why can’t OpenAI provide the best idea?

[lshaan Gupta @ishaan80]

@alexandr_wang Why do we still need prompts? Should we just give it a role and @Muse should auto select the best available prompt/skill to do the job?

[lshaan Gupta @ishaan80]

So cool!

[NVIDIA AI @NVIDIAAI]

When several people talk at once, a transcript can get messy fast.

Our new Nemotron 3 Diarization model tracks who spoke when, even when voices overlap. It handles up to eight speakers, has 100M parameters, and is now available on @huggingface 🤗

[lshaan Gupta @ishaan80]

@louislecat I don’t open my inbox any more so why does cleaning it matters?

[lshaan Gupta @ishaan80]

Interesting update to the Robotics world!

[Black Forest Labs @bfl_ai]

Introducing FLUX 3 Action.

An open weights 7B World Action Model that achieves first place on the RoboLab benchmark.

It outperforms the previous best open model by 6.1 percentage points while using 56% fewer parameters and running up to 3.95x faster.⁠⁠

FLUX 3 Action removes the usual trade-off between world action model performance and VLA speed: it still predicts video and actions together, but plans more than twice as far ahead and runs faster per second of robot motion than the strongest open VLA.

Teams can fine-tune FLUX 3 Action on their own demonstrations to create policies for a particular robot and task. Together with @nvidia, we also integrated FLUX 3 Action natively into @huggingface's LeRobot, with fine-tuning recipes included and edge deployment on NVIDIA Jetson.

Beyond robotics, we’re also seeing promising results training task-specific policies for acting in simulated environments like gaming, controlling a vehicle, computer use, and wherever else a model needs to understand a visual environment and then choose what to do next.

FLUX 3 Action builds on the same image, video, and audio pretraining as FLUX 3, but uses a smaller architecture designed for practical deployment. In midtraining, we trained the model to predict actions and future frames together.

We’re releasing the weights, code, fine-tuning recipe, benchmarks, and reproducible examples so researchers and developers can build on the model with their own robots, environments, and tasks (see below).

[lshaan Gupta @ishaan80]

This sounds interesting for people with ADHD symptoms to track their body clock.

[Anish Moonka @anishmoonka]

Up to 78 percent of people with ADHD have a body clock that runs late, and moving that clock earlier has been linked to fewer ADHD symptoms. A 2025 paper by two Canadian doctors pulled the evidence together and laid out a treatment plan.

Melatonin, the hormone that tells your body it's time to sleep, starts rising about 45 minutes later at night in children with ADHD and about 90 minutes later in adults. Brandon Luu of Queen's University and Nicholas Fabiano of the University of Ottawa gathered this research in Frontiers in Psychiatry. Sleep problems affect up to 80 percent of adults with ADHD and up to 82 percent of children.

People with ADHD also tend to have a flatter, later daily cortisol rhythm, and some studies have found a smaller pineal gland, the tiny gland deep in the brain that makes melatonin. Their body clock genes show weaker daily swings too.

Melatonin and bright light therapy have both moved the body clock earlier in children and adults with ADHD. Early studies link that shift to fewer ADHD symptoms, and sleep programs for children have improved symptoms, sleep, and daily functioning.

Screening every ADHD patient for sleep and body clock problems is the first step in the plan the authors propose. After that come a fixed wake-up time, bright light in the morning, less light and screen time at night, regular daily routines, and low-dose melatonin for people whose clock is clearly running late.

The paper is a perspective piece that summarizes earlier research, and much of the treatment evidence is still early. For the large group of people with ADHD whose clocks run late, the first steps it recommends are about light, timing, and routine. https://t.co/hmKclkvAUH

[lshaan Gupta @ishaan80]

@bryan_johnson So years of research and data boils down to just good sleep?

[lshaan Gupta @ishaan80]

@DarioAmodei @chandrarsrikant Sounds more like AI controlling humans and telling them which experiments to run, with humans like guinea pigs doing that job. So when the robots arrive, they can replace humans.

[lshaan Gupta @ishaan80]

@jeffreykWNDR Valid point! With the advent of modern technology, our way of working evolves. New skills need to be acquired and things that were never possible before will seem normal. Human productivity will increase like never before.

[lshaan Gupta @ishaan80]

@patmatthews Building in the space of Physical AI

[Bain Capital Ventures @BainCapVC]

Make money. Have fun. Live with integrity. Everything else can change.

Fund XI. $1.6B total capital for those who know that it will.

[lshaan Gupta @ishaan80]

@atShruti Can one attend online?

[lshaan Gupta @ishaan80]

@FinMinIndia @nsitharamanoffc @DFS_India @NPCI_NPCI @DDNewslive @airnewsalerts @PIB_India @MIB_India @CBC_MIB Countries Singapore charge MDR for paynow transaction. It is essential to implement MDR or all the tax payers end up paying it.

[lshaan Gupta @ishaan80]

@INCIndia Merchants pay MDR in Singapore as well for transactions beyond a certain value on their paynow network. What’s wrong in this?

Eventually this money will be paid by all tax payers if not charged from merchants. Would you like to apply for MS in finance from Harvard?

[lshaan Gupta @ishaan80]

@naval Or agents can go rogue and kill humans so that they donot need to get the job done.

[lshaan Gupta @ishaan80]

This is interesting but will it live up to the hype? How about the harness? Will it be as good as ChatGPT and Claude?

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution

[lshaan Gupta @ishaan80]

@satyanadella Very valid point.

[Wellness at Work AI @luminaforwork]

Thank You @designnominees! Happy to see Lumina as App of the Day.

[lshaan Gupta @ishaan80]

@MiteshKhapra @ai4bharat Which train is the Steel Express? It sounds natural but how is the semantic understanding to correct sentence formation?