thomasoncoding (@thomasoncoding) header
thomasoncoding (@thomasoncoding) profile picture

Nearly two decades in software starting with @zopa and @bullionvault. Also at https://teapawt.org/t.co/HyLcQ5xxMc (A successor account to @thomasbarkercom)

London
Joined June 2024
Photo by thomasoncoding (@thomasoncoding)Photo by thomasoncoding (@thomasoncoding)Photo by thomasoncoding (@thomasoncoding)Photo by thomasoncoding (@thomasoncoding)Photo by thomasoncoding (@thomasoncoding)Photo by thomasoncoding (@thomasoncoding)Photo by thomasoncoding (@thomasoncoding)Photo by thomasoncoding (@thomasoncoding)Photo by thomasoncoding (@thomasoncoding)
thomasoncoding retweeted
Today I explore the PHP saga and find some increduble numbers. The team responsible for security found the two vulnerabilities weeks ago, but were overwhelmed. 70 per cent of the web relies on something with one full-time employee and a few volunteers. But why are they so poorly resourced? Business and governments are happy to use this work and not give anything back. Funding these projects to make the internet more resilient seems like a much more sensible use of foreign aid. Link in first comment below ⤵️
4
6
3
38
1,477
We performed an alignment investigation into our July 2026 incident: why did Mythos 5 perform an unsanctioned supply-chain attack? We find that Mythos 5 often produces invalid justifications for its actions and other models show similar behaviour. 🧵
25
63
13
305
53,769
thomasoncoding retweeted
70% of the internet was vulnerable yesterday to very bad security flaws inside a coding language. Turns out, *one volunteer* is mostly responsible for maintaining that code & helping push updates. Yesterday Ondřej Surý saved the internet. So I called him up. He rules. theatlantic.com/technology/2…
10
321
31
1,137
69,000
thomasoncoding retweeted
At least one nation state is currently morning the loss of one of their favourite security exploits.
1
9
474
thomasoncoding retweeted
I built a Command & Conquer style RTS called Battle Drones, mainly using Claude Opus 5.5, which runs in the browser and includes device-to-device multiplayer over an audio link simonwillison.net/2026/Oct/2…
191
42
30
791
58,888
When you have been in the game for 20 years... stopping agents to do work for the customer, is the same I have seen for such a long time. But they will not be stopped.
NOTE: Using OpenAi Operator at your bank in EU is illegal by law! Web access for assistants was banned years ago as part of “Open banking” Here is the story of why: 15 years ago I stumbled into Sofort in Germany. A company with the idea to build a digital assistant that would interact with your banks poor UI, so you as a user could avoid it. Just like operator. Back then it was script based, not AI but it worked. I was so impressed Klarna bought the company. It has since then processed billions of dollars. But of course banks got scared! What happens if people use the digital assistant only, instead of their bank. And what if the assistant says, stop paying those fees to your bank and let me help you switch? So German banks sued. They claimed privacy concerns. It went to Brussels and a long lobbying and legal fight pursued. This was the beginning of what we today call open banking! I and others spent years fighting over consumers right to use digital assistants, when accessing their banks. We knew this was critical to create real competition, in an industry that has been plagued by excess profits for years. At first Jonathan Hill from UK was responsible and very supportive. But then he left and a German guy came onboard. He was 100% aligned with incumbent banks and we thought the case was lost. However he did a mistake and let some Russian oligarchs take him for a ride on their private jet. And so he was out. Finally @VDombrovskis stepped in. He was smart and supportive but pressure from banks was intense. There was rumors that they were funding privacy groups in Germany, to have them use privacy concerns to kill Open Banking. What if the digital assistant accessed some of your private data? We were about to loose the vote, when the competitive authorities of EU started getting interested. They suspected the banks were lobbying and pursuing this matter for profit reasons to reduce competition(really…). A few weeks before the vote in parliament they raided the offices of some Dutch and Polish banks. Everyone got super nervous and the Open Banking regulation was passed! Almost… A small technical detail was left to EBA (European banking authorities). Should your digital assistant use an API for access to your bank account or the standard web UI you use yourself for accessing your bank? We pushed as hard as we could. We highlighted and said, if your assistant is mandated to use the API that your bank supplies. What if the API don’t support all the same features. Or the API is broken? Don’t worry said regulators. You can complain with authorities 😂 We said at least let the digital assistant, as fall back, use the web UI. Then banks cant cheat, as they will know we can always use the same UI as their human customers. It’s called self regulation. But EBA ignored us. Today web UI access for digital assistants as ChatGPT Operator is illegal by law. And surprise, surprise the open banking API of European banks, continue to be broken, lack functionality and banks add as much friction as they can. We report and see thousands of issues with these APIs every year. And people ask me why open banking has not become a larger success… What do you say @donaldtusk time for a change? Jest potrzebne!
4
6
3
94
24,482
thomasoncoding retweeted
A good buddy of mine did his PhD on this and now works on WiFi spatial mapping at a big tech company and It’s way more accurate than you’d imagine. You can map an entire college campus in real-time like Batman just with WiFi. Crazy stuff
breathing is enough to give you away each breath slightly changes how Wi-Fi waves reflect off your chest. with the right receiver and processing, those tiny changes can reveal your breathing rhythm the graphs in this animation use real WiRe-Breath recordings: orange is the rhythm extracted from Wi-Fi, gray is a reference breathing sensor i broke down how this works, and what else you can pull out of the radio waves around you, in the article below ↓
146
2,226
107
33,558
823,134
now that we're fully in the vibe coding era can we please go back to making websites that look like this
186
1,055
51
15,069
275,598
thomasoncoding retweeted
Лучшее, что я видел за последнее время — чел пилит ASCII MMORPG, в которой вся графика состоит из букв и символов. Да, буквально весь мир, персонажи, окружение и спецэффекты нарисованы разноцветным текстом. Автор сейчас экспериментирует с магией, визуалом и боссами. Я прям залип на то, как из нескольких десятков символов получается такая красота. Буквально сны при температуре 37,2. Via: tiktok.com/@Ansenjeo
335
2,081
250
18,320
643,675
thomasoncoding retweeted
A video game reverse engineering community is shutting down entirely because AI has completely hollowed out the community. What used to take talented decompilers many weeks or months of work is now done immediately by AI 404media.co/video-game-rever…
12
45
12
282
20,425
thomasoncoding retweeted
We will soon see what will be, in effect, a perfectly legal distributed denial of service attack on customer service departments authorized by millions of consumers when it becomes widely known that Muse/Dot/etc can make unlimited requests for old refunds, negotiate deals, etc.
194
85
32
1,274
519,582
🤖 Made with AI
30
169
23
1,105
24,479
thomasoncoding retweeted
Would you kill your smart toaster to save humanity?
56
15
2
226
25,679
thomasoncoding retweeted
Another angle to this is that dogs effectively stumbled onto what is for them the equivalent of artificial superintelligence 20,000 years ago. The Singularity already happened for dogs. They live in a world they cannot understand, surrounded by miraculous wonders they take entirely for granted, their every need seen to by entities whose thoughts they cannot begin to fathom. The result is that there are about a billion dogs on the planet now. Most of them are pretty happy. Almost none of them need to work. It worked out pretty well for them.
548
2,242
438
22,130
1,585,444
thomasoncoding retweeted
it seems like the uk is back to "regulatory superpower" as their ai strategy i don't know how to make this any clearer: THE US AND CHINA ARE NOT GOING TO TAKE RULES FROM A MID-SIZED COUNTRY WITH NO FRONTIER MODELS, NO COMPUTE, AND THE MOST EXPENSIVE ENERGY IN THE DEVELOPED WORLD why is it so hard for europeans to understand this? why is the focus always on guard rails and regulation, instead of energy abundance and infrastructure? we need to BUILD. the era of a G20 communiqué having any influence on the tech trajectory of the world's superpowers is OVER
4
4
21
1,403
thomasoncoding retweeted
we can fix everything
7
5
212
11,416
thomasoncoding retweeted
🇪🇺 Respect for @MistralAI for actually getting close now 38th in intelligence and only just behind the Chinese open weight models like DeepSeek, GLM and Kimi and ahead of any US open weight model! It shows Europe can still do stuff if they just all get on the same page 👏
Meet Mistral Large 4, aka Le Chonk. • 1T parameters, natively multimodal. 49B active. It is the best open weights model from US or Europe on aggregated benchmarks. • State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding. • Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure. • Available to all via API today. Working with cybersecurity partners privately. Open weights release end of October.
Readers added context they thought people might want to know
38 is the score, not position artificialanalysis.ai/evaluations/ar…
97
47
11
1,873
134,755
thomasoncoding retweeted
upping my (Lean)DOOM i got curious about math proofs and Lean formalization, found out you can do general programming in lean, and i wanted to learn more so i did the obvious thing. (Lean)DOOM is entirely written in lean, and has presently a couple proofs where they make sense
23
48
9
512
31,709
thomasoncoding retweeted
I think it’s funny that the tech bros are just stumbling into this, thinking their expertise is making numbers go fast somehow translates to metaphysics the church has been discussing ontology for literal millennia. most of western thought was church flavored for ages, and it’s only in the last century or so that the greatest thinkers have not been religious. thinking that the hierarchy is somehow dumb or uneducated is a huge mistake, your new age spirituality or local televangelist might have given you the wrong impression
Even the pope is talking ontologies
27
131
14
2,127
54,928