Jainam Parmar
@jainamparmar.bsky.social
AI Newsletter Pioneer at The Shift | Get the latest breakthroughs, tools, and strategies in AI | Subscribe to my newsletter 👇🏻
https://theshiftai.beehiiv.com/subscribe
https://theshiftai.beehiiv.com/subscribe
AI is not going to take job.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
The Shift
Get the latest breakthroughs, tools, and strategies to level up at your work all in under 5 minutes a day, delivered 5 days a week.
theshiftai.beehiiv.com
January 28, 2026 at 2:39 PM
AI is not going to take job.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
5/ But the security tradeoff is huge.
- Give it full computer access and one malicious DM could mean prompt injection, leaked credentials, hijacked workflows.
- Give it full computer access and one malicious DM could mean prompt injection, leaked credentials, hijacked workflows.
January 28, 2026 at 2:39 PM
5/ But the security tradeoff is huge.
- Give it full computer access and one malicious DM could mean prompt injection, leaked credentials, hijacked workflows.
- Give it full computer access and one malicious DM could mean prompt injection, leaked credentials, hijacked workflows.
4/ It’s also insanely flexible.
- You can route the “brain” through OpenAI, Anthropic, or -Google.
- People are already wiring it into Notion, Todoist, daily audio recaps.
- You can route the “brain” through OpenAI, Anthropic, or -Google.
- People are already wiring it into Notion, Todoist, daily audio recaps.
January 28, 2026 at 2:39 PM
4/ It’s also insanely flexible.
- You can route the “brain” through OpenAI, Anthropic, or -Google.
- People are already wiring it into Notion, Todoist, daily audio recaps.
- You can route the “brain” through OpenAI, Anthropic, or -Google.
- People are already wiring it into Notion, Todoist, daily audio recaps.
3/ You can control it through WhatsApp, Telegram, Signal, Discord, even iMessage.
- Calendar updates, form fills, email actions, daily admin.
- Agents start feeling… normal.
- Calendar updates, form fills, email actions, daily admin.
- Agents start feeling… normal.
January 28, 2026 at 2:39 PM
3/ You can control it through WhatsApp, Telegram, Signal, Discord, even iMessage.
- Calendar updates, form fills, email actions, daily admin.
- Agents start feeling… normal.
- Calendar updates, form fills, email actions, daily admin.
- Agents start feeling… normal.
2/ What makes people freak out is simple.
- Moltbot runs on your own device and does tasks, not just text.
- A DM becomes a command center.
- Moltbot runs on your own device and does tasks, not just text.
- A DM becomes a command center.
January 28, 2026 at 2:39 PM
2/ What makes people freak out is simple.
- Moltbot runs on your own device and does tasks, not just text.
- A DM becomes a command center.
- Moltbot runs on your own device and does tasks, not just text.
- A DM becomes a command center.
1/ It’s a local AI agent built by Austrian dev Peter Steinberger.
Launched in December as “Clawdbot,” renamed after Anthropic trademark pressure.
Now it’s going viral fast.
Launched in December as “Clawdbot,” renamed after Anthropic trademark pressure.
Now it’s going viral fast.
January 28, 2026 at 2:39 PM
1/ It’s a local AI agent built by Austrian dev Peter Steinberger.
Launched in December as “Clawdbot,” renamed after Anthropic trademark pressure.
Now it’s going viral fast.
Launched in December as “Clawdbot,” renamed after Anthropic trademark pressure.
Now it’s going viral fast.
If Earth-2 really outperforms a state-of-the-art system released just months ago, this isn’t linear progress.
It’s compounding.
Long-range weather prediction may be shifting from uncertainty to strategic advantage faster than anyone expected.
It’s compounding.
Long-range weather prediction may be shifting from uncertainty to strategic advantage faster than anyone expected.
January 27, 2026 at 2:03 PM
If Earth-2 really outperforms a state-of-the-art system released just months ago, this isn’t linear progress.
It’s compounding.
Long-range weather prediction may be shifting from uncertainty to strategic advantage faster than anyone expected.
It’s compounding.
Long-range weather prediction may be shifting from uncertainty to strategic advantage faster than anyone expected.
This matters most for regions without massive forecasting infrastructure.
- Governments that relied on delayed or inconsistent models suddenly get near-instant signals.
- Weather stops being guesswork and starts becoming operational intelligence.
- Governments that relied on delayed or inconsistent models suddenly get near-instant signals.
- Weather stops being guesswork and starts becoming operational intelligence.
January 27, 2026 at 2:03 PM
This matters most for regions without massive forecasting infrastructure.
- Governments that relied on delayed or inconsistent models suddenly get near-instant signals.
- Weather stops being guesswork and starts becoming operational intelligence.
- Governments that relied on delayed or inconsistent models suddenly get near-instant signals.
- Weather stops being guesswork and starts becoming operational intelligence.
Their Global Data Assimilation model rebuilds a full global weather snapshot fast.
And the Nowcasting model zooms in on the next 0–6 hours using satellite data.
That means real-time storm awareness, anywhere with coverage.
And the Nowcasting model zooms in on the next 0–6 hours using satellite data.
That means real-time storm awareness, anywhere with coverage.
January 27, 2026 at 2:03 PM
Their Global Data Assimilation model rebuilds a full global weather snapshot fast.
And the Nowcasting model zooms in on the next 0–6 hours using satellite data.
That means real-time storm awareness, anywhere with coverage.
And the Nowcasting model zooms in on the next 0–6 hours using satellite data.
That means real-time storm awareness, anywhere with coverage.
That alone is wild.
- But the bigger shift is speed.
- Nvidia says forecasts that used to burn half of a supercomputer’s capacity can now run in minutes on GPUs.
- But the bigger shift is speed.
- Nvidia says forecasts that used to burn half of a supercomputer’s capacity can now run in minutes on GPUs.
January 27, 2026 at 2:03 PM
That alone is wild.
- But the bigger shift is speed.
- Nvidia says forecasts that used to burn half of a supercomputer’s capacity can now run in minutes on GPUs.
- But the bigger shift is speed.
- Nvidia says forecasts that used to burn half of a supercomputer’s capacity can now run in minutes on GPUs.
Nvidia’s big bet is Earth-2.
Instead of physics-heavy simulations, it uses transformer-based AI.
They claim it beats DeepMind’s GenCast on 70+ forecasting variables, up to 15 days out.
Instead of physics-heavy simulations, it uses transformer-based AI.
They claim it beats DeepMind’s GenCast on 70+ forecasting variables, up to 15 days out.
January 27, 2026 at 2:03 PM
Nvidia’s big bet is Earth-2.
Instead of physics-heavy simulations, it uses transformer-based AI.
They claim it beats DeepMind’s GenCast on 70+ forecasting variables, up to 15 days out.
Instead of physics-heavy simulations, it uses transformer-based AI.
They claim it beats DeepMind’s GenCast on 70+ forecasting variables, up to 15 days out.
AI is not going to take job.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
The Shift
Get the latest breakthroughs, tools, and strategies to level up at your work all in under 5 minutes a day, delivered 5 days a week.
theshiftai.beehiiv.com
January 7, 2026 at 2:36 PM
AI is not going to take job.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
3/ The bigger signal is strategic.
- Nvidia is assembling an ecosystem, not shipping parts.
- If this stack becomes the default, Nvidia could play the same foundational role for physical machines that Android plays for smartphones.
- Nvidia is assembling an ecosystem, not shipping parts.
- If this stack becomes the default, Nvidia could play the same foundational role for physical machines that Android plays for smartphones.
January 7, 2026 at 2:36 PM
3/ The bigger signal is strategic.
- Nvidia is assembling an ecosystem, not shipping parts.
- If this stack becomes the default, Nvidia could play the same foundational role for physical machines that Android plays for smartphones.
- Nvidia is assembling an ecosystem, not shipping parts.
- If this stack becomes the default, Nvidia could play the same foundational role for physical machines that Android plays for smartphones.
2/ On the compute side, Nvidia introduced Rubin, its next architecture after Blackwell.
- The company claims 3.5× faster training and 5× faster inference, raising the ceiling for large-scale AI workloads.
- The company claims 3.5× faster training and 5× faster inference, raising the ceiling for large-scale AI workloads.
January 7, 2026 at 2:36 PM
2/ On the compute side, Nvidia introduced Rubin, its next architecture after Blackwell.
- The company claims 3.5× faster training and 5× faster inference, raising the ceiling for large-scale AI workloads.
- The company claims 3.5× faster training and 5× faster inference, raising the ceiling for large-scale AI workloads.
1/ At the core is a generalist robotics stack.
- Open models like Cosmos and Isaac GR00T let robots reason, plan, and act across many tasks, while tools like Isaac Lab-Arena and OSMO standardize training and deployment.
- Open models like Cosmos and Isaac GR00T let robots reason, plan, and act across many tasks, while tools like Isaac Lab-Arena and OSMO standardize training and deployment.
January 7, 2026 at 2:36 PM
1/ At the core is a generalist robotics stack.
- Open models like Cosmos and Isaac GR00T let robots reason, plan, and act across many tasks, while tools like Isaac Lab-Arena and OSMO standardize training and deployment.
- Open models like Cosmos and Isaac GR00T let robots reason, plan, and act across many tasks, while tools like Isaac Lab-Arena and OSMO standardize training and deployment.
AI is not going to take job.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
The Shift
Get the latest breakthroughs, tools, and strategies to level up at your work all in under 5 minutes a day, delivered 5 days a week.
theshiftai.beehiiv.com
January 6, 2026 at 2:04 PM
AI is not going to take job.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
3/ This creates real tension.
- As reliance grows, OpenAI faces lawsuits over mental health concerns and is calling for clearer FDA pathways.
- Healthcare is already one of AI’s biggest real-world use cases, and the next phase will be about balancing access, and safety.
- As reliance grows, OpenAI faces lawsuits over mental health concerns and is calling for clearer FDA pathways.
- Healthcare is already one of AI’s biggest real-world use cases, and the next phase will be about balancing access, and safety.
January 6, 2026 at 2:04 PM
3/ This creates real tension.
- As reliance grows, OpenAI faces lawsuits over mental health concerns and is calling for clearer FDA pathways.
- Healthcare is already one of AI’s biggest real-world use cases, and the next phase will be about balancing access, and safety.
- As reliance grows, OpenAI faces lawsuits over mental health concerns and is calling for clearer FDA pathways.
- Healthcare is already one of AI’s biggest real-world use cases, and the next phase will be about balancing access, and safety.
2/ A huge share of usage isn’t clinical, it’s administrative.
- ChatGPT now handles 1.6 to 1.9 million insurance-related questions every week.
- People use it to decode bills, compare plans, and prepare disputes, essentially outsourcing the most confusing parts of healthcare.
- ChatGPT now handles 1.6 to 1.9 million insurance-related questions every week.
- People use it to decode bills, compare plans, and prepare disputes, essentially outsourcing the most confusing parts of healthcare.
January 6, 2026 at 2:04 PM
2/ A huge share of usage isn’t clinical, it’s administrative.
- ChatGPT now handles 1.6 to 1.9 million insurance-related questions every week.
- People use it to decode bills, compare plans, and prepare disputes, essentially outsourcing the most confusing parts of healthcare.
- ChatGPT now handles 1.6 to 1.9 million insurance-related questions every week.
- People use it to decode bills, compare plans, and prepare disputes, essentially outsourcing the most confusing parts of healthcare.
1/ Health already makes up more than 5 percent of all ChatGPT messages.
- Nearly half of users use it to understand medical terms, and 44 percent use it to explore treatment options.
- Around 70 percent of these conversations happen outside clinic hours, when real-world support isn’t available.
- Nearly half of users use it to understand medical terms, and 44 percent use it to explore treatment options.
- Around 70 percent of these conversations happen outside clinic hours, when real-world support isn’t available.
January 6, 2026 at 2:04 PM
1/ Health already makes up more than 5 percent of all ChatGPT messages.
- Nearly half of users use it to understand medical terms, and 44 percent use it to explore treatment options.
- Around 70 percent of these conversations happen outside clinic hours, when real-world support isn’t available.
- Nearly half of users use it to understand medical terms, and 44 percent use it to explore treatment options.
- Around 70 percent of these conversations happen outside clinic hours, when real-world support isn’t available.
AI is not going to take job.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
The Shift
Get the latest breakthroughs, tools, and strategies to level up at your work all in under 5 minutes a day, delivered 5 days a week.
theshiftai.beehiiv.com
December 30, 2025 at 2:39 PM
AI is not going to take job.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
Our newsletter, The Shift, delivers breakthroughs, tools, and strategies to help you become value creator and build in this new era easily.
Subscribe: theshiftai.beehiiv.com/subscribe
Plus: Get access to 3k+ AI Tools and free AI courses when you join.
3/ The results back it up.
- Self-play SWE-RL delivers 7 to 10 point gains on SWE-bench Verified and Pro, beating models trained on human data.
- More importantly, it points to coding agents that can keep improving on their own, using existing codebases as infinite training ground.
- Self-play SWE-RL delivers 7 to 10 point gains on SWE-bench Verified and Pro, beating models trained on human data.
- More importantly, it points to coding agents that can keep improving on their own, using existing codebases as infinite training ground.
December 30, 2025 at 2:39 PM
3/ The results back it up.
- Self-play SWE-RL delivers 7 to 10 point gains on SWE-bench Verified and Pro, beating models trained on human data.
- More importantly, it points to coding agents that can keep improving on their own, using existing codebases as infinite training ground.
- Self-play SWE-RL delivers 7 to 10 point gains on SWE-bench Verified and Pro, beating models trained on human data.
- More importantly, it points to coding agents that can keep improving on their own, using existing codebases as infinite training ground.