Startup Operator, Voice AI ; ex-founder ; MBA'27 at @LBS
London
- Tweets 188
- Following 71
- Followers 11
- Likes 18
Pinned Tweet
New here.
Building voice AI in London. Ran AI transformation at a global energy group before this; MBA'27 at LBS.
Posting what building and selling AI looks like - numbers, customers, opinions.
Building with AI, especially in London? Say hi. Who should I be following?
3
123
1/ everyone's comparing muse, chatgpt and claude on smarts. i think consumer agents get won on something duller: who clears the most annoying admin from your week without you checking its work
1
5
4/ the catch is trust. i think the winner is whoever gets good at knowing when to check with you first. too many asks and it's useless, one bad send in your name and it's switched off for good
1
3
deepseek went from the cheap training story to a $75b valuation in under two years. the bit i keep coming back to is $1b run rate while most of their models are free to download. open weights clearly didn't stop them making money
JUST IN: 🇨🇳 DeepSeek raises $7.5 billion at $75 billion valuation; annualized revenue hits $1 billion, more than doubling in months - The Information.
1
20
muse at $100 a month only works if it clears real admin. the demo that stuck with me was it booking 3d printer time for the team. if it can do that for my parking tickets and gym cancellations, i'd pay. if it's a smarter chatbot, most people won't
JPM says "Muse has the potential to become the most widely used consumer AI application since ChatGPT. "
Is that true, or just the case until the next shinier, faster agent pops up in 5-7 days?
1
emulate incorporated in august and is reportedly raising $700m at $3.7bn with no product. london is quietly where ex-deepmind people now raise the biggest seed rounds on earth
1
8
screenshot via silicon uk, citing the FT: silicon.co.uk/e-enterprise/s…
2
openai is now asking for international standards around recursive self-improvement. fair. but a standard only matters if labs report the same incidents before the scary demo becomes a press release.
2
14
I think humanoid robots are being judged on the wrong thing. A perfect demo tells you almost nothing about whether a factory should buy one.
Only 7,000 were sold worldwide last year. The gap between attention and actual deployment is enormous.
1
2
32
Factories already know which jobs hurt: understaffed shifts, unsafe motions, the same lift 4,000 times a day. Start there.
A robot that handles one ugly task every day is worth more than one that can perform twenty tricks for a camera.
1
4
Anthropic may launch a new model before its IPO. That should make people more sceptical of the next benchmark chart, not less. When a lab is selling intelligence and shares at the same time, capability claims become investor relations.
1
13
Temporal just raised $550m at a $12.55bn valuation. The company makes sure long-running software can resume after something fails.
My bet: reliability becomes a bigger AI category than memory.
1
2
25
Temporal says OpenAI's usage grew 60x in under a year. That makes sense: the more tools an agent can touch, the more ways it can break halfway through a job.
Model capability creates demand for recovery infrastructure.
1
16