Jason Gorman's Avatar

Jason Gorman

@jasongorman.bsky.social

If you're serious about your investment in your software development capability, visit https://www.codemanship.co.uk

2,603 Followers  |  230 Following  |  7,852 Posts  |  Joined: 07.02.2024
Posts Following

Posts by Jason Gorman (@jasongorman.bsky.social)

We know he doesn't write them. We know he rarely even reads them. Ask yourself who this power would be handed down to. The techbros' POTUS-in-waiting?

28.02.2026 22:14 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

The mainstream news media has no playbook for politicians answering questions directly. TBF, it's been a long time since any politician has.

28.02.2026 21:48 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

"It would be a huge distraction"

U-huh

28.02.2026 21:33 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Bear with me on this. What if owning stuff isn't a skill, and people shouldn't get paid for it?

28.02.2026 21:04 β€” πŸ‘ 10    πŸ” 1    πŸ’¬ 0    πŸ“Œ 0
Video thumbnail

BREAKING β€” 51 Iranian children are dead after a strike hits Minab girls elementary school in Iran.

28.02.2026 12:33 β€” πŸ‘ 2970    πŸ” 1941    πŸ’¬ 148    πŸ“Œ 703

Corporate sales: people who don't understand what they're buying negotiating with people who don't understand what they're selling.

28.02.2026 19:24 β€” πŸ‘ 7    πŸ” 1    πŸ’¬ 0    πŸ“Œ 0

The most realistic part of the show is potential customers negotiating with people who have zero understanding of what they're selling.

28.02.2026 19:15 β€” πŸ‘ 8    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

I learn so much about business from The Apprentice.

You design a product following the direction of the loudest person in the room.

And then, after the product can't be changed, you do market research.

28.02.2026 19:09 β€” πŸ‘ 11    πŸ” 4    πŸ’¬ 1    πŸ“Œ 0
Preview
People power led the Greens to victory in Gorton and Denton The party’s historic byelection win was earned by listening to voters, not ignoring them, as Labour has done

The Green Party took on the billionaire donors.

And won.

Some of my thoughts πŸ‘‡πŸΌ

observer.co.uk/news/opinion...

28.02.2026 18:34 β€” πŸ‘ 558    πŸ” 120    πŸ’¬ 9    πŸ“Œ 7
A white female plumber, a Jewish gay man and a Muslim wan walked into a hall....and united a community.

A white female plumber, a Jewish gay man and a Muslim wan walked into a hall....and united a community.

Join.greenparty.org.uk

28.02.2026 18:35 β€” πŸ‘ 2168    πŸ” 414    πŸ’¬ 41    πŸ“Œ 22

How can I explain royalty to people in countries that don't have hereditary monarchies?

We tried selective breeding of humans, and it didn't work.

28.02.2026 17:48 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

And in that sense, it's not a product focus. It's a problem focus.

28.02.2026 15:11 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Every requirement's a hypothesis, and every release is an experiment to test that hypothesis.

If we add a "Best Vegan Takeaways" bar to the home page, do sales of vegan takeaway go up? Do sales of non-vegan takeaway go down?

It helps to have clear, testable goals.

28.02.2026 15:11 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

"Ah, but what is 'accuracy'?"

I refer the honourable gentleman to my previous answers about people who resort to philosophy in discussions of practical matters.

28.02.2026 15:08 β€” πŸ‘ 6    πŸ” 1    πŸ’¬ 1    πŸ“Œ 0

For teams deploying multiple times a day, this has always been true

28.02.2026 14:56 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

This blows the business models of the hyper-scalers out of the water. (Not that they actually had a business model, but let's pretend that losing money on every token is a plan.)

So who'll train the hyper-scale "teacher" models? I think research facilities like CERN may provide a bit of a clue.

28.02.2026 10:30 β€” πŸ‘ 5    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

And it would run on a high-end gaming GPU.

I think I'd probably bet on that as the more likely future.

28.02.2026 10:28 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

Hyper-scale LLM inference may never be economically viable, for a whole host of reasons.

But a hyper-scale LLM can be distilled into an SLM that can get within the ballpark of it's "teacher's" performance at a fraction of the inference cost - we're talking orders of magnitude.

28.02.2026 10:28 β€” πŸ‘ 5    πŸ” 0    πŸ’¬ 1    πŸ“Œ 2

Just flipping through the Rolodex in my mind of "What if" scenarios relating to strategies built around LLMs - e.g., what if the hyper-scalers go bust? - and how the risks could be mitigated.

The only Plan B that would work in *every* scenario is "people".

28.02.2026 09:04 β€” πŸ‘ 5    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Going forward, perhaps it should be "First, there are moral and ethical issues"?

28.02.2026 08:57 β€” πŸ‘ 6    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

CNN reporting that the strikes on Iran tonight are joint US and Israeli strikes.

Which means Donald Trump should be impeached for a third time.

What he has done tonight is illegal under the US constitution and the UN charter.

28.02.2026 07:16 β€” πŸ‘ 7950    πŸ” 2339    πŸ’¬ 252    πŸ“Œ 122

Trump had no choice but to act. The Iranians were just weeks away from obtaining the missing Epstein files.

28.02.2026 07:45 β€” πŸ‘ 1537    πŸ” 364    πŸ’¬ 38    πŸ“Œ 21
Preview
How Can We Progress Past β€œAI” Woo? For decades, when I’ve interacted with folks working in machine learning, one of the big questions that comes up is β€œHow do we test these systems?” Traditional software’s be…

Our social media feeds are overflowing with all kinds of claims about "AI" coding assistants and "agents". It's hard to see a clear signal in all the noise.

What's really real and what really works?

We're not going to figure it out with "feels".

codemanship.wordpress.com/2026/02/28/h...

28.02.2026 07:50 β€” πŸ‘ 5    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Epstein files contain explicit but unsubstantiated claim that Trump abused minor Department of Justice did not release FBI memos when it uploaded millions of pages of files beginning in December

The Guardian obtained the missing 302s. This story destroys the DOJ's excuse that they withheld the last three interviews because they were duplicative. The accuser covered different ground in different interviews.

27.02.2026 21:44 β€” πŸ‘ 1247    πŸ” 559    πŸ’¬ 28    πŸ“Œ 21

With a current cash burn rate of $billions, how long might Anthropic be able to limp on without their main investors?

28.02.2026 05:21 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Or with a debate.

28.02.2026 05:06 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

I'm interested in what's real and what works. You don't discover that with a committee.

28.02.2026 05:03 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

That's a shame. I'm really hoping to find more and better data. When we're testing hypotheses about ML systems, we really need to look for trends.

28.02.2026 04:47 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

Can we see the session logs?

28.02.2026 03:41 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

People? Just a thought.

28.02.2026 03:39 β€” πŸ‘ 3    πŸ” 1    πŸ’¬ 0    πŸ“Œ 0