Every AI drop, the minute it ships ⚡ New models, tools, agents & prompts, explained fast. Daily AI news + how-to.
Joined January 2009
- Tweets1.3K
- Following2.5K
- Followers1K
- Likes2.7K
"It's given me purpose and a reason to fight for every second I'm on this earth."
Zach McManus (@DeadManRolln) lives with ALS and says he's Neuralink's 22nd participant. Asked about regrets, he said: "there is no negative, none." His advice anyway: do your research.
What it does for him so far:
• Phone, laptop, TV, Xbox and a robot vacuum by thought: NOW REAL for him, in trial
• Driving his wheelchair by thought: still ahead for him (per his Oct 31 post)
• Slowing ALS itself: not what the implant does
Don't get me wrong: "no negatives" is one man's experience. It's brain surgery with an investigational device, and Neuralink's first participant had threads retract in 2024 before software changes recovered performance. But someone with ALS describing a reason to fight is hard to dismiss.
If you were weighing an implant, what would you trust more: other participants' stories, or published long-term safety data?
Neuralink Patient 22 shares his regrets on getting his brain computer interface implant.
I sat down with @DeadManRolln and he shared his incredible story.
Watch on JoshWest247 YT 📺
Link Below: ⬇️
3
Most proposals don't lose. They expire.
In Proposify's data on 742,137 proposals, won deals closed an average of 2.5 days after the buyer opened them. "Follow up in a week" runs on the wrong clock.
The free 2.5-Day Follow-Up Clock (ours, untested):
Day 0: send within a day of the call
Day 1-2: opened? Ask "Which option is closest?"
Day 3: not opened? Resend with a 2-line summary
Any day: changes asked? Reply the same day
Decision date: "Hold your spot or close this out?"
Don't get me wrong: it's one vendor's own customers, a pattern, not proof. But one-person shops in that data closed 37% vs 33% for the biggest teams.
5-line proposal rule, AI prompt and the money math in the article 👇
Your first fix: A) sign and date every proposal, or B) chase the ones sitting in "maybe"?
742,137 proposals: the ones signed before sending saw a 65% higher close rate, and won deals closed 2.5 days after the buyer opened them. A free AI proposal + follow-up kit 👇
5
742,137 proposals: the ones signed before sending saw a 65% higher close rate, and won deals closed 2.5 days after the buyer opened them. A free AI proposal + follow-up kit 👇
Take a train through Sweden's far north and your video call usually dies somewhere between cell towers.
Norrtåg ran Starlink alongside the mobile networks on its trains. Its Nov 9 results:
• +78% average internet capacity on Umeå–Luleå
• +58% on Storlien–Sundsvall–Umeå
• Combined capacity above 1,000 Mbps in some tests
Already in regular service on several trains. 5 more being fitted now; whole fleet during 2027 (Norrtåg's stated plan).
Don't get me wrong: that's capacity to the train, shared by everyone aboard, not per passenger. And Norrtåg hasn't said what it costs.
Still, for people who commute through Norrland, a working connection on board turns travel hours into usable hours.
My bet: if these numbers hold through winter, other rural rail lines copy it.
Would reliable Wi-Fi make you pick the train over driving?
NEWS: Sweden’s Norrtåg is expanding Starlink to every train. 🇸🇪
• Starlink is already operating on several trains, with the entire fleet set to be equipped during 2027.
• Adding Starlink increased average internet capacity by 78% on the Umeå–Luleå route.
• Tests on the Storlien–Sundsvall–Umeå route showed a 58% increase.
• Combined Starlink and mobile network capacity exceeded 1,000 Mbps in some tests, shared across the train.
• The results were so encouraging that Norrtåg accelerated the rollout. Five more trains are currently being equipped.
• For passengers, that means better connections for work, video calls and streaming through areas where mobile coverage struggles.
Another win for Starlink: bringing better internet to people travelling through Sweden’s remote north.
21
"To contribute to research for other people with conditions like mine."
That's half of why Paul, a man with MND, became Neuralink's first patient in the UK. The other half: his own freedom.
His story, labeled honestly:
• Cursor by thought the day after surgery, then home: NOW REAL (UCLH)
• Using it at home with daily practice: NOW REAL (UCLH)
• Favourite games like Dawn of War: IN TRIAL (being explored with engineers)
• Approval outside studies, or slowing MND: STILL A GOAL
Not all good: brain surgery, an investigational device, and no published UK results yet. Not all bad: months later he wrote, "It's an incredible piece of technology that has changed my life!"
Why he said yes, and what he wanted back first 👇
What would you want back first: A) the game you love, or B) messages, bills and browsing without asking for help?
Neuralink's first UK patient called his nerves about brain surgery "a huge understatement." The day after, he went home moving a cursor by thought. Why he said yes 👇
17
Neuralink's first UK patient called his nerves about brain surgery "a huge understatement." The day after, he went home moving a cursor by thought. Why he said yes 👇
Fable 5.2 costs twice what Opus 5.5 does per token ($8/$40 vs $4/$20 per million). Anthropic's own guidance still says start most work on Opus 5.5.
If you sell AI-built work at a fixed price, that gap is your margin.
The fix order from Anthropic's Nov 9 model guide, before you pay for the bigger model:
1) Give it a way to check its own work (a test, a build, a script).
2) Raise effort one level.
3) Only then move up a model, starting at a lower effort.
AI Edge's breakdown adds a "finish the task" prompt for when Fable 5.2 stops early. Worth pasting into your system prompt.
Not all good: these are Anthropic's own benchmark runs, not independently reproduced. And bigger isn't always pricier: on one coding test, Fable 5.2 at high matched Opus 5.5 at max for roughly half the cost per task.
OUR TAKE: picking the right model for each job is quietly becoming a billable skill. Test on 5 of your real jobs before you pick a default.
Which do you reach for first: the cheaper model or the strongest one?
30
Crash a motorbike deep in Japan's mountains and your phone usually says no service. One Osaka rescue crew now drives the signal to you.
Portable Starlink, labeled honestly:
1. RESCUE-1: LIVE. Starlink Mini on its trucks, free Wi-Fi for riders at no-signal crash sites
2. Swedish sailor off South Africa: DONE. A Starlink call got him a tow before the wind hit (Oct 26)
3. Mini 2: NOT SHIPPING YET. Up to 4.5 h on its battery, per SpaceX
4. Price: $999 first offer incl. a year of service (reported)
5. One mountain recovery: about 10 hours
Don't get me wrong, a dish isn't an emergency service. But a hurt rider calling home from a ravine is real.
The full story is below 👇
Crash with no signal: a Mini 2 in your own bag, or a crew that brings the Wi-Fi?
Starlink Mini 2 runs 4.5 hours on a battery. A rescue crew in Japan already shows who portable Starlink is really for: the rider stuck in a ravine with no signal 👇
25
Starlink Mini 2 runs 4.5 hours on a battery. A rescue crew in Japan already shows who portable Starlink is really for: the rider stuck in a ravine with no signal 👇
“How fast can you type with your thumbs?”
Ex-Tesla president Jon McNeill says that question, asked by @elonmusk in the back of a Model S in Hong Kong, is where Neuralink started. Elon's numbers in the story: about 18 words a minute on a phone vs about 60 on a keyboard. Then: “What if you could think into the machine?”
Where that question stands today:
• Text and voice by thought for people with ALS: NOW REAL in trial participants. Per Neuralink, one reached 75 words a minute in their first week (5.8% error rate, 125K-word vocabulary)
• Thought input faster than speech, for anyone: STILL A GOAL
Don't get me wrong: it's one man's memory of one car ride, and early data from one participant in an investigational trial, not peer-reviewed. But a complaint about thumb speed becoming a voice for people who lost theirs is quite a path.
What should brain-to-text unlock first: talking with family again, or getting back to work?
Ex-Tesla President Jon McNeill reveals Elon Musk's idea for Neuralink started in the back of a Model S in Hong Kong, with Elon asking how fast he could type.
Elon Musk then asked for the top brain-computer interface researchers’ papers on the spot, and met them the next week:
“Elon and I were traveling together to China. We just landed in Hong Kong. We’re in the back of a Model S, and we’re both catching up on our inbox on our phones. And he looks over at me and says, how fast can you type with your thumbs?”
“And I said, I have no idea. I’ve never timed it. He’s like, well, the average human’s 18 words a minute. He said, how fast can you type on a keyboard? I said, I think it’s like 60 or 70 words a minute. So let’s call it 60.”
“He said, exactly. So you are more than three times faster on the keyboard than you are on your phone. This device has slowed the input into the machine. And he’s like, that’s a big deal, because you can’t get speed of answers and decision making if you’re going three times slower.”
“He’s like, what if you could think into the machine? I was like, well, that would blow this constraint away, because now we don’t have keyboards and thumbs to go through. You’re thinking in real time into the machine. He’s like, exactly. That would pull the front-end constraint off of the compute for humans. That’s a big deal.”
“So he didn’t sit on these ideas. He’s very unique in this way. And he said to his chief of staff, Sam Teller, who’s in the front seat: Sam, can you track down the top neuro computer science interface researchers in the world? Get their papers.”
“I want to read the papers while we’re here in China and on the flight back, and I want to meet with these people next week, because I want to start to work on brain-computer interface. And just like that, fast.”
“So by the next week, he’s meeting with some of the top minds in the world on this topic, and he doesn’t let go until he starts Neuralink.”
25
Apple Stores have started scanning returned iPhones through the sealed box, per Bloomberg's Mark Gurman. People were returning boxes that looked new, minus the phone.
A small shop that refunds the moment a carrier scans the label is trusting every parcel.
The free 5-Check Return Rule, for refunds above your own threshold:
1. Refund fires after check-in, not on the first scan
2. Item, size and serial match the order
3. Tags on, nothing missing
4. Three photos on arrival
5. Check the customer's 90-day pattern, hold and review, never accuse
Don't get me wrong: most returns are honest, and the 9% fraud figure comes from big retailers and a returns vendor. That's why the rule targets the few, not everyone.
Return Log, AI prompt and the worked math in the article 👇
Your first fix: A) when your refund fires, or B) a log of repeat returners?
Retailers say 9% of returns are fraud, and Apple Stores have started scanning returned iPhones through the sealed box. A free 5-check return kit for small online shops 👇
12
Retailers say 9% of returns are fraud, and Apple Stores have started scanning returned iPhones through the sealed box. A free 5-check return kit for small online shops 👇
"Internet at school is a necessity. Don't delay it any longer, let alone with just promises."
A Ternate city councillor in Indonesia said that on Oct 2 about schools on the city's outer islands (Habartimur, translated).
On Nov 10, this report said the city has now installed Starlink at 14 schools on Moti, Hiri and Batang Dua, prioritizing the worst-connected ones. The video shows a Starlink box being handed over.
Why it matters there:
• Teachers needed it for teaching work and school admin, the councillor said.
• Until spring 2025, the health centre on Batang Dua had no internet at all. Its head said staff got online by finding locals who ran a business with Starlink (Tribun Ternate).
Don't get me wrong: a dish isn't a lesson. These schools still need power, devices, trained teachers and someone paying every month, and the 14-school count comes from this report; I haven't seen the city's own list yet. That clinic's own link came from Telkomsat, not Starlink, so there's more than one way to close the gap.
My bet, if it holds: the first visible change is teacher training and admin, not kids streaming.
What should these 14 schools use it for first: teacher training or student lessons?
BREAKING: Starlink connects 14 schools across three remote islands in Indonesia. 🇮🇩
Ternate’s city government has installed Starlink at schools on Moti, Hiri and Batang Dua, prioritizing those struggling with poor internet access.
1
29
In 2011, a woman paralyzed by a stroke nearly 15 years earlier drank coffee using a robot arm she steered by thought.
Now Neuralink has opened a trial to stroke survivors. Up to 10 people. One catch: the stroke has to have missed the spot where the implant goes.
Labeled honestly:
• Stroke survivors driving robot arms and speech with implants: NOW REAL IN RESEARCH (not Neuralink)
• Cursor at about mouse speed: NOW REAL in Neuralink trial participants
• Neuralink in stroke survivors: IN TRIAL (PRIME2, recruiting)
• Moving their own arm again: STILL A GOAL
Not all good: brain surgery, 10 people, safety first. Not all bad: a real hospital and a named surgeon are now recruiting.
Full breakdown, and the fine print 👇
Which would matter more after a stroke: A) a phone and computer by thought, or B) waiting for tech that moves the hand itself?
Neuralink's first stroke trial has room for 10 people, and the stroke has to have missed the spot where the implant goes. Here's what stroke survivors have already done with brain implants 👇
17
Neuralink's first stroke trial has room for 10 people, and the stroke has to have missed the spot where the implant goes. Here's what stroke survivors have already done with brain implants 👇
Miles Deutscher runs a 50+ person media business. He says he doesn't even ask creatives for a resume.
His playbook: hire from X. Look for people who consistently post great work in your niche (ideally 3K-15K followers), judge taste over credentials, then run a 1-2 week paid trial.
Flip it if you're the creative. A 10-minute profile check tonight:
1) Pin your single best finished piece of work.
2) Make your niche obvious in the first line of your bio.
3) Post one finished piece this week, not a thread about plans.
4) Before any trial, agree in writing that it's paid and what "done" looks like.
Not all good: this is one hiring manager's method, and follower count isn't skill. Plenty of strong creatives post rarely. Still, consistent public work is the part you control today.
OUR TAKE: as AI makes average content cheap, visible taste becomes the hireable skill.
Which would you rather be judged on: your resume or your last 10 posts?
Every tech business in the world is looking for creative/content talent.
Yet, nobody can hire them.
As someone with a 50+ person media business, THIS is how you hire your creative talent:
1. Hire from 𝕏
The best creatives are already posting. Look for people who consistently put out great work in your niche, and reach out to them directly (ideally 3-15k followers).
The talent density on 𝕏 is insane. Grok Bot can even help you here.
You can run the same outreach strategy on small Instagrammers, LinkedIn writers, etc.
2. Judge creativity and taste (not resumes)
I've never hired from a top school. In fact, I don't even ask for resumes.
Creatives' online presence should be their resume.
Hunt for: work ethic, attention to detail, video/written presentation skills over fancy fluff metrics.
3. Run a paid trial
Always run a 1-2 week paid trial before fully committing.
4. Invest in their development
Pay agencies to train them, work closely with your creatives to build taste, etc.
It's totally unrealistic to expect someone to come in and know everything. Even if they've had a content/creative job before, the likelihood that the responsibilities are the same as yours is very low.
24
A sailor finally gets a signal mid-ocean. Their kid starts talking. And they're watching a data counter.
Starlink at sea, labeled honestly:
1. Asahi Tanker: LIVE. Starlink on 103 ships
2. Doctor by video + AED on every ship: IN ROLLOUT, done by Jan 2027
3. Ship-shore data sharing: STATED PLAN
4. Seafarers' connectivity score worldwide: FELL, 7.46 to 6.89 (data caps)
5. Maritime data blocks: up 50% from Dec 1 (reported)
Don't get me wrong, a dependable call home from a tanker is real progress. Complaints about caps mean the signal itself is no longer the question.
The full story is below 👇
Your fleet has one more dollar per crew member: bigger data allowances, or video doctor visits?
Starlink now lets crews on 103 Japanese ships call home from the middle of the sea. But a global seafarer survey just found the internet score falling, and the reason is the next fight at sea 👇
20
Starlink now lets crews on 103 Japanese ships call home from the middle of the sea. But a global seafarer survey just found the internet score falling, and the reason is the next fight at sea 👇
Alex Finn doesn't want AI to help him do more work. He wants the same work with more head space.
His prompt asks your agent to review what it knows about you and suggest systems, subagents or routines that "lighten my cognitive load." 700+ people bookmarked it in its first two hours.
Before you build anything it suggests, run each idea through 3 checks:
1) Does it happen every week? If not, skip it.
2) What breaks if it's wrong? Anything that sends, pays or deletes stays draft-only until you approve it.
3) Can you check its work in under a minute? If not, it adds load instead of removing it.
Not all good: the answer is only as good as what your agent remembers. A fresh chat will guess, so give it your real weekly tasks first. "Saved me more time than any other prompt" is his experience. Results vary.
Still, it's free, copy-paste and usable tonight.
OUR TAKE: the skill that pays next is deciding which work an agent should carry, not doing more of it yourself.
Which would you hand off first: email triage or weekly reporting?
You need to run the prompt below in Grok Bot/Duet/Dots
It's saved me more time than any other single prompt I've run
The biggest issue with agents is you use it to 1% of it's capability. You only automate 1% of the work it can actually take care of
This fixes that
Prompt:
My goal for working with you is to lighten my cognitive load. I want to get more work done, but with less effort on my part. Based on what you know about me and what we do together, what are some things we can do, systems we can implement, subagents we can build, or routines we can create to accomplish this goal?
This will cause your agent to review everything it knows about you and the work you've done together, and come up with systems that 'lighten your cognitive load'
For me this is the biggest benefit of AI
I don't even care about getting more work done. I just want to get the same amount of work done, but have more head space I can spend on things I enjoy
Everything I do with AI is in an effort to reach this north star
Use this prompt and watch the magic happen (feel free to just send this post to your agent)
14
Penn Medicine asked patients why they'd missed an appointment. The top answer: "I did not know I had an appointment."
In two NHS trials, one line in the reminder cut no-shows from 11.1% to 8.4%: the specific cost of the missed slot. A vague version did worse.
The free 5-Line Reminder Rule:
1. Who + when: name, day, date, time
2. One specific number about the slot
3. One-reply move: YES or MOVE
4. A clear change-by time
5. A human exit for anything sensitive
Don't get me wrong: this is healthcare research adapted for small businesses, not a tested small-business result. But it costs nothing to try, and you can count the difference in 30 days.
No-Show Sheet, AI prompt and the worked math in the article 👇
First fix: A) your reminder wording, or B) a personal call to your repeat no-shows?
One line changed in a reminder text cut missed appointments from 11.1% to 8.4% in two randomized trials. A free no-show kit for any business with a calendar 👇
1
1
14