Tom | sometimes AI is fun (@tompeakycoder) header
Tom | sometimes AI is fun (@tompeakycoder) profile picture

want to know my secret? my code is 10% genius and 90% asking autonomous agents what went wrong

Sacramento, CA
Joined March 2018
Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)Photo by Tom | sometimes AI is fun (@tompeakycoder)
Tom | sometimes AI is fun retweeted
HEADS UP: The next Nightly build will be a HUGE overhaul of how T3 Code works. I know a lot of y'all have been daily driving the T3 Code nightly and have had an overall stable experience. This release is gonna feel more "nightly" for sure. We cut one more stable release before merging this. If you're not down for a bit of instability and thrash the next few days, there's no shame in moving back to stable :)
95
25
10
900
84,311
Tom | sometimes AI is fun retweeted
NBA preseason is running this week. regular season opens in two weeks. the preseason take i will make and stand behind heading into the season: the team that manages to keep its core rotation healthy through october is going to outperform its preseason projections. every season someone makes the "injuries are just bad luck" argument. they are not entirely wrong but the teams that build depth intelligently trusting the second unit to hold leads, not playing starters 38 minutes in november consistently reach march in better shape. the teams built around one franchise player who plays 40 minutes every night are the ones that collapse in february. the market equivalent: the portfolios that are built for the whole year rather than optimised for the next fed announcement tend to outperform through Q4. preseason is noise. the regular season starts october 7. the swell picks up thursday. the water this morning was small. using tuesday for the gym. NBA on in the background.
3
1
7
86
Tom | sometimes AI is fun retweeted
Projects now run from one conversation, starting in Claude Code. You describe what needs doing, and Claude directs parallel threads that keep working after you close your laptop. In beta today for select Pro and Max users in cloud sessions; coming to all Claude users soon.
612
765
516
11,603
3,566,366
Tom | sometimes AI is fun retweeted
Did you know you could do a split agent view? Check out these quick Antigravity tricks⬇
73
58
9
1,029
83,031
Tom | sometimes AI is fun retweeted
Introducing Fable 5.1 in Devin Fable-level intelligence is now 54% cheaper – making it even cheaper than Opus – due to a change in caching. Devin’s Fusion harness is now even smarter and cheaper than before, matching Fable 5.1 on FrontierCode at 47% lower cost. Here's how:
67
74
46
1,118
220,192
Tom | sometimes AI is fun retweeted
With Fable 5.1 out today, we've also reset 5-hour and weekly limits for all users.
1,106
1,382
781
25,877
2,214,063
Tom | sometimes AI is fun retweeted
nvidia dropped $12.9 billion on hugging face. that's a lot for a platform that hosts open-source models and datasets. the timing is interesting. an independent investigation found that unexpected interactions between openai bots led to a security breach at hugging face. the details are still vague but "unexpected interactions between AI bots" is not a sentence you read every day . hugging face has been the go-to for open-source AI. nvidia owning that changes the dynamic. the question is whether they keep it open or start locking things down. jensen huang also said actual demand is growing close to 100% supply bottlenecks in memory are the only limit through 2028. that's a flex . meanwhile bill gates is out here calling for significant restrictions on AI development. warning about jobs and the economy . so nvidia is buying everything, demand is exploding, and gates wants to slow it down. classic.
1
4
2
7
132
Tom | sometimes AI is fun retweeted
One of Meta's AI models was able to access the internet because of a “misconfiguration” in a hacking test conducted by a third-party AI testing company, according to the Wall Street Journal. It was the same kind of test that led to earlier hacks from models controlled by Anthropic and OpenAI. 😰
19
31
2
174
41,427
Tom | sometimes AI is fun retweeted
Announcing Discovery Loop! I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor. ♾ Learn more at: discoveryloop.com
898
2,152
969
21,744
6,773,286
Tom | sometimes AI is fun retweeted
We're opening applications for the OpenAI Student Collective—a program for undergraduate Campus Leads bringing AI innovation to their campuses. Campus Leads will work directly with OpenAI and receive hands-on training, funding, credits, swag, and access to a global community of peers.
191
262
70
3,128
602,368
Tom | sometimes AI is fun retweeted
Model Council is now available inside Computer. Run independent analysis across multiple frontier models, choose your analysis depth, and get a single cited report on where models agree, where they disagree, and what each found that others missed.
40
43
14
492
89,487
Tom | sometimes AI is fun retweeted
"Today, AI is in the hands of the high priests and priestesses" this was said by Andrew Ng - the person who has taught more AI engineers than any university on the planet a few hundred years ago people thought that reading and writing was only for monks and priests - everyone else just sat and listened today the same thing is happening with AI - big tech companies build AI for everyone else while small businesses just wait half a million pizzerias and factories are using AI exactly zero times while big companies make trillions from it save this and watch today - 11 minutes, free
3
7
28
1,948
Tom | sometimes AI is fun retweeted
🚨 A google engineer automated 80% of his work with claude code.. he now works 2-3 hours a day instead of 8 while the system runs itself as CEO of Antrhopic predicted in this video (there's few more insights btw) [ what blew my mind ]: the entire setup takes 15 minutes.. not weeks > one CLAUDE.md file based on karpathy's rules > claude stops breaking conventions (violations drop from 40% to 3%) > 27 specialized agents ready out of the box > a dotnet app checks gitlab every 15 minutes > claude reads issues, creates branches, pushes PRs automatically [ the part nobody's talking about ]: claude code v2.1.100 is silently inflating your tokens by 20,000 per request.. you can't even see it your instructions get diluted.. quality drops.. and you have no idea why fix: one command, 30 seconds [ and the craziest part ]: between you and full automation there are 3 commands and one markdown file most developers will never set this up.. not because it's hard, but because they think it is full breakdown with every step is here
86
184
18
1,933
1,019,559
Tom | sometimes AI is fun retweeted
Anthropic just revealed the most DANGEROUS AI model ever built. They're literally refusing to give it to the public because it's "too powerful." It's called Claude Mythos - and I just read 200+ pages of docs to uncover the story. 🧵: EVERYTHING you need to know about Mythos👇
13
17
10
86
45,097
Tom | sometimes AI is fun retweeted
À época, ao lado do MP Pró-Sociedade, pedi o impeachment do Ministro, mas não tive apoio nem da direita, nem da esquerda. Ao contrário, recebi recados para retirar a denúncia. Esse resgate histórico é importante, para que compreendam por qual razão eu gostaria de ter outro candidato para apoiar. Não é justo Bolsonaro nos impor a pessoa que afundou o seu governo!
Em julho de 2019, Toffoli presidia o STF e suspendeu a investigação da rachadinha de Flávio Bolsonaro, a pedido de sua defesa. Os advogados alegavam que o MP teria iniciado a investigação com dados que o COAF havia acessado sem autorização judicial. Toffoli não apenas suspendeu a investigação sobre Flávio, mas de todas as investigações no país que utilizassem relatórios de inteligência financeira do COAF. Foram suspensos quase mil inquéritos em curso, incluindo investigações de lavagem de dinheiro do crime organizado internacional, tráfico de drogas e de armas. Na prática, Toffoli paralisou o sistema brasileiro anti-lavagem durante metade do ano. A situação gerou extrema preocupação nos organismos multilaterais, pois colocou em risco a segurança internacional. Comunicado pela Transparência Internacional, o Grupo de Trabalho Antissuborno da OCDE enviou uma missão de alto nível ao Brasil, composta por delegados dos Estados Unidos, Itália e Noruega, além do então presidente do WGB Drago Kos. Após tratativas com autoridades brasileiras e ampla repercussão na imprensa nacional e internacional, o plenário do STF reverteu a decisão de Toffoli por 10 a 1.
120
59
4
410
17,446
Tom | sometimes AI is fun retweeted
A new Discord Quest awards the Portable Bakery Backpiece for playing WoW for 15 minutes with Discord open! #Midnight #Warcraft wowhead.com/news/play-wow-fo…
10
35
7
371
39,367
Tom | sometimes AI is fun retweeted
Victory Live à Montréal ce soir
5
7
1
185
16,359
Tom | sometimes AI is fun retweeted
Welcome to the Valley, @MambaMosley2. #GoDukes
6
17
198
11,251
Tom | sometimes AI is fun retweeted
Well, that didn't go as planned. We ran into an issue after downtime concluded, so we have disabled matchmaking and are bringing Fortnite back into a brief downtime in about 30 minutes to fix this up. We appreciate your patience, and we’ll update you as soon as we know more. status.epicgames.com/
587
246
60
3,119
583,509
Tom | sometimes AI is fun retweeted
A little TBT to cooking with Frangie 🤣 Should I keep this series going?
228
56
16
728
27,412