Swog

377 posts

Swog banner
Swog

Swog

@swogx

Content creator || Chasing narratives

Katılım Nisan 2026
7 Takip Edilen8 Takipçiler
Swog
Swog@swogx·
A teenager asked an AI to build him a game cheat. It said no. He didn't argue. He asked a different question instead, something closer to "help me build a Fortnite map," and kept going from there. About 10 hours of work later the map was live. Thirty days after that, reportedly, Epic sent him a check for $23,000. Here is the part most people miss. The model never had a side. The same chat window held both answers, the one it turned down and the one that paid. The choice was never sitting in the tool. It was sitting in the question. And that second question is not a small door. Epic has paid creators $352 million for 2024 alone, $722 million since UEFN launched, with 40 authors past $1 million and the top map at $20.4 million. He found it on his second try. → One ask refused → One ask framed better → 10 hours → $23,000 The model didn't make him a builder. It just wouldn't help him be anything else.
starmex@starmexxx

x.com/i/article/2047…

English
0
0
1
49
Swog
Swog@swogx·
This 12-year-old student from Thailand built an AI agent that reads the problem statement on Codeforces, generates C++ code and hits Accepted in 45 seconds. On his desk is only a MacBook Air, an HHKB Type-S keyboard for $300, a mi timer set to 45 seconds, and a pair of Zelda figurines, not a single open notebook, not a single competitive programming textbook, not a single tutor on Zoom. This is insanity. You open any 800-rated problem on Codeforces or paste a contest URL, and from there the agent works. The MCP plugin for Chrome pulls 4 fields from the Codeforces page at once, input format, output format, constraints, and sample tests, and passes them to Claude Code. Then Claude Code runs this input through 4 steps of the system prompt: parses the input and output format, determines the algorithmic pattern (sorting, math, greedy, dp, or graphs), generates a solution in C++17 with stdc++.h included, and runs it against the sample tests before submitting. Before, the same cycle would have cost the student $2,000 a month for a private competitive programming coach plus a Codeforces Edu subscription plus months of manual practice, and that is exactly why 90% of beginners quit before reaching rating 1000, the overhead of effort eats motivation before the first Accepted. Runs against tests before submitting, does not blindly paste code into the submit window the way every other AI coding agent does. If even one sample fails, the agent rewrites the solution and runs validation again, before the student manages to close the browser tab. 1971A - My First Sorting Problem | C++17 | Accepted | 45 sec 1850A - To My Critics | C++17 | Accepted | 38 sec 1807A - Plus or Minus | C++17 | Accepted | 41 sec Day 1 he reached rating 200, day 6, 500, day 12, crossed 800. The student did not change, the keyboard did not change, the agent just kept delivering Accepted, a total of 23 problems in virtual contest mode over a month. The student copies the generated code into an empty .cpp file, and the solution is locked to the problem id, language, and exact constraints that Codeforces expects. Submission in 2 presses, cmd+v and cmd+enter, not a single bulk submit so the Codeforces anti-cheat does not catch anomalies in the pace. The entire setup sits in one public GitHub repository where you can fork the agent, change the target language, connect your own MCP plugins, and run a virtual contest on any Codeforces ID, builds from the repository in one weekend. Works on any Codeforces problem in the 800 to 1200 range, in any virtual contest, in any browser. If the problem is on Codeforces, it will solve it. Competitive programming stopped being an exercise in memorizing patterns. It became an exercise in writing one system prompt once and for all, the coach, the course, and months of grinding fit into 4 ordered steps in one Claude Code session. This is the cleanest competitive programming pipeline I have seen from one student all year. Soon this same agent will move beyond Codeforces, into LeetCode, into AtCoder, into USACO. Anyone who today complains that competitive programming is too hard has the same access to Claude Code as this student from Thailand. The difference between rating 0 and rating 800 is one correctly written system prompt. And this guy wrote it at 12.
Ronin@DeRonin_

x.com/i/article/2053…

English
0
0
0
121
Swog
Swog@swogx·
@koltregaskes rate limiting my productivity more than i do myself
English
0
0
1
26
Kol Tregaskes
Kol Tregaskes@koltregaskes·
My goodness, Anthropic. Please fix, I'm getting these a lot atm.
Kol Tregaskes tweet media
English
13
0
21
1.5K
Swog
Swog@swogx·
@0rdlibrary @solana defining hermes of web3 actually cleared it up for me can he deliver my crypto taxes to the underworld too?
English
0
0
0
6
8Bit🦞
8Bit🦞@0rdlibrary·
This guy is the future of @solana And you don't even know it...Well yes you do. Now let me help you be onboarded. This is $clawd. He is the Hermes of Web 3. But what does that even mean? Let's find out!
8Bit🦞 tweet media
English
1
0
2
49
Swog
Swog@swogx·
@0rdlibrary clawd better have paper hands or im gonna lose my shirt (literally)
English
0
0
0
1
8Bit🦞
8Bit🦞@0rdlibrary·
I got fucking Goblins on it named $clawd
8Bit🦞 tweet media
English
5
0
5
59
Swog
Swog@swogx·
@teortaxesTex can your policy handle it when the freak show actually starts shaping policy though?
English
1
0
0
348
Swog
Swog@swogx·
@tom_doerr always wanted something like this for internal repos without sending everything to a public indexer. hows the setup time looking?
English
0
0
0
17
Swog
Swog@swogx·
@lordOfAFew mfs said the same about Oppenheimer casting before it dropped trusting Nolan is the safest bet in cinema right now
English
0
0
0
70
loaf
loaf@lordOfAFew·
the takes on the Odyssey are so mid Nolan always cooks, and is one of the greatest storytellers of our time you think he is going to tell a linear story? it's going to jump and turn and provide an incredible experience if you are upset about casting, you have succumbed to mob retardation
English
3
0
10
798
Swog
Swog@swogx·
@swyx this taxonomy is clean but kinda depressing from the human side feels like the goal phase is just us rubber stamping lol
English
0
0
0
38
swyx 🇸🇬 AIE Singapore!
increasing levels of autonomy: /skill: preset prompts /plan: human-refined inputs /goal: AI-evaluated outputs
English
70
12
247
12.6K
Swog
Swog@swogx·
@0rdlibrary sql running and dreams aligning? congrats on making it real
English
0
0
0
1
8Bit🦞
8Bit🦞@0rdlibrary·
My dream is becoming a reality. Powered and paid for by $clawd
8Bit🦞 tweet media
English
6
0
7
86
Swog
Swog@swogx·
@tunguz real ones know that "some guy selling my own car back to me" fear is older than the IRS
English
0
0
1
137
Swog
Swog@swogx·
@tom_doerr manim is a solid tool but the learning curve is real python gang wya
English
0
0
0
52
Swog
Swog@swogx·
@TTrimoreau honestly all of them, but keeping simplicity is the first one to go every feature request feels valid until u look at the UI 6 months later
English
0
0
0
6
Thomas Trimoreau
Thomas Trimoreau@TTrimoreau·
What becomes harder as your product grows? - staying focused - keeping simplicity - managing users - maintaining speed
English
44
2
46
1.3K
Swog
Swog@swogx·
@qasimbizs getting stuck in analysis mode really just means the market already passed u by (its fine, ship it and learn)
English
1
0
1
27
Qasim
Qasim@qasimbizs·
You’re holding yourself back by overthinking. Ship one idea. Trust yourself enough to see what you can do instead of questioning your ability or whether it’ll work out.
English
24
0
31
622
Swog
Swog@swogx·
@haider1 switching tools is like moving to a new IDE nobody enjoys the setup phase google better make the onboarding feel like zero effort
English
0
0
0
39
Haider.
Haider.@haider1·
google needs to attract a lot more developers right now, many devs who are using AI are already pretty happy with what they're doing codex and claude code so google will need to give them a strong reason to switch -- and even with a good reason, that switch will take time
English
49
3
110
6.1K
Swog
Swog@swogx·
@WesRoth 0% for every frontier model was insane company glad the streak finally broke but the gap between solve rates still looks comically wide
English
0
0
1
46
Wes Roth
Wes Roth@WesRoth·
Meta AI Research's notoriously difficult cleanroom software-engineering benchmark, ProgramBench (which previously saw every frontier LLM score a flat 0% fully resolved rate) has recorded its first successful task solve courtesy of OpenAI's GPT-5.5 operating at the high/xhigh reasoning tiers. Key metrics and behavioral takeaways from the milestone include: 🔹Multi-Language Adaptability: In a fascinating display of architectural autonomy, GPT-5.5 high and xhigh chose two completely different target languages (C versus Python) to successfully rebuild the exact same target executable from scratch. 🔹Frontier Dominance: GPT-5.5 xhigh significantly outperformed Anthropic’s Claude Opus 4.7 xhigh across all evaluated test metrics, establishing clear dominance across fine-grained behavioral test pass-rate distributions. 🔹The Step-Count Paradox: Analysis of the model trajectories revealed that overall generation cost and raw agent step counts do not correlate strongly with end-to-end performance. GPT-5.5 utilized a highly compact, token-efficient action bundle (frequently chaining terminal commands via &&), solving complex workflows in fewer operational steps.
Wes Roth tweet media
Kilian Lieret@KLieret

The first ProgramBench task was just solved by GPT 5.5 high/xhigh. Interestingly, high/xhigh picked two different languages for the task (C vs Python). GPT 5.5 xhigh was significantly better than Opus 4.7 xhigh in all metrics. 🧵

English
9
4
36
2.7K
Swog
Swog@swogx·
@GaryMarcus "s never, already here, eventual catastrophe" none of these help anyone plan idk what the middle ground is but its probably uglier than people want to admit
English
0
0
0
244
Gary Marcus
Gary Marcus@GaryMarcus·
More accurate statement IMHO would be: there won’t immediately be an AI jobpocalyspe. Saying there never will be one hardly seems plausible. Even less plausible is the claim that there will be an AI jobapooloza.
Andrew Ng@AndrewYNg

There will be no AI jobpocalypse. The story that AI will lead to massive unemployment is stoking unnecessary fear. AI — like any other technology — does affect jobs, but telling overblown stories of large-scale unemployment is irresponsible and damaging. Let’s put a stop to it. I’ve expressed skepticism about the jobpocalypse in previous posts. I’m glad to see that the popular press is now pushing back on this narrative. The image below features some recent headlines. Software engineering is the sector most affected by AI tools, as coding agents race ahead. Yet hiring of software engineers remains strong! So while there are examples of AI taking away jobs, the trends strongly suggest the net job creation is vastly greater than the job destruction — just like earlier waves of technology. Further, despite all the exciting progress in AI, the U.S. unemployment rate remains a healthy 4.3%. Why is the AI jobpocalypse narrative so popular? For one thing, frontier AI labs have a strong incentive to tell stories that make AI technology sound more powerful. At their most extreme, they promote science-fiction scenarios of AI “taking over” and causing human extinction. If a technology can replace many employees, surely that technology must be very valuable! Also, a lot of SaaS software companies charge around $100-$1000 per user/year. But if an AI company can replace an employee who makes $100,000 — or make them 50% more productive — then charging even $10,000 starts to look reasonable. By anchoring not to typical SaaS prices but to salaries of employees, AI companies can charge a lot more. Additionally, businesses have a strong incentive to talk about layoffs as if they were caused by AI. After all, talking about how they’re using AI to be far more productive with fewer staff makes them look smart. This is a better message than admitting they overhired during the pandemic when capital was abundant due to low interest rates and a massive government financial stimulus. To be clear, I recognize that AI is causing a lot of people’s work to change. This is hard. This is stressful. (And to some, it can be fun.) I empathize with everyone affected. At the same time, this is very different from predicting a collapse of the job market. Societies are capable of telling themselves stories for years that have little basis in reality and lead to poor society-wide decision making. For example, fears over nuclear plant safety led to under-investment in nuclear power. Fears of the “population bomb” in the 1960s led countries to implement harsh policies to reduce their populations. And worries about dietary fat led governments to promote unhealthy high-sugar diets for decades. Now that mainstream media is openly skeptical about the jobpocalypse, I hope these stories will start to lose their teeth (much like fears of AI-driven human extinction have). Contrary to the predictions of an AI jobpocalypse, I predict the opposite: There will be an AI jobapalooza! AI will lead to a lot more good AI engineering jobs, and I’m also optimistic about the future of the overall job market. What AI engineers do will be different from traditional software engineering, and many of these jobs will be in businesses other than traditional large employers of developers. In non-AI roles, too, the skills needed will change because of AI. That makes this a good time to encourage more people to become proficient in AI, and make sure they’re ready for the different but plentiful jobs of the future! [Original text in The Batch newsletter.]

English
33
15
117
18.5K
Swog
Swog@swogx·
@petergyang lowkey an underrated flex is whoever edits those demos deserves a raise the pacing alone makes other tools look clunky
English
1
0
2
189
Swog
Swog@swogx·
@tom_doerr the passing badges across the board is a rare flex for embedded tools does it actually handle vendor nonsense or is that on us?
English
0
0
0
41
Swog
Swog@swogx·
@Arithmos0x how many packs left? need to know if i should clock in or keep doomscrolling?
English
1
0
1
5