π€ The newest AIs can now write another AI by hand. Older ones score close to zero
On Monday, MIT researchers showed AI models editing another model's weights directly, with no training. Alex Krentsel, a Berkeley PhD researcher, reran one of their tests on older and newer models.
Each model had to write all the numbers inside a small network that plays the Atari game Breakout, in a single reply, with no training and no tools. The yardstick is a network trained by trial and error for 500 million steps.
GPT-6 Astra's network scored 22 times what that yardstick got. Claude Opus 5.5 hit 20x, GPT-6.1 Sol 18x. GPT-5 to GPT-5.5 and Opus 4.5 to 4.8 didn't even reach the yardstick.
Krentsel calls it "a new, step-change emerging capability that's appeared in just the latest set of models."
His own caveat: the network is tiny, 20,486 parameters, while frontier models have around a trillion. On top open-weight models it hasn't worked so far.
π t.me/techpsyche
On Monday, MIT researchers showed AI models editing another model's weights directly, with no training. Alex Krentsel, a Berkeley PhD researcher, reran one of their tests on older and newer models.
Each model had to write all the numbers inside a small network that plays the Atari game Breakout, in a single reply, with no training and no tools. The yardstick is a network trained by trial and error for 500 million steps.
GPT-6 Astra's network scored 22 times what that yardstick got. Claude Opus 5.5 hit 20x, GPT-6.1 Sol 18x. GPT-5 to GPT-5.5 and Opus 4.5 to 4.8 didn't even reach the yardstick.
Krentsel calls it "a new, step-change emerging capability that's appeared in just the latest set of models."
His own caveat: the network is tiny, 20,486 parameters, while frontier models have around a trillion. On top open-weight models it hasn't worked so far.
π t.me/techpsyche
Your Ego is Your Biggest Bug
If youβre the smartest person in the room, youβre in the wrong room. Donβt get defensive during code reviews.
A critique of your code is not a critique of your soul. Welcome the feedback.
π t.me/techpsyche
If youβre the smartest person in the room, youβre in the wrong room. Donβt get defensive during code reviews.
A critique of your code is not a critique of your soul. Welcome the feedback.
π t.me/techpsyche
PAKISTANπ΅π° Funprime Technology is HiringJunior Backend Developer
π€Organization: Vinspyre Pvt. Ltd.
πLocation: I-11/3, Islamabad - Pakistan
π’ Job Type: Onsite
πΌ Experience: 1β2 Years
Required Skills:
* Hands-on experience with Python & FastAPI
* Knowledge of REST APIs and backend development
* Experience with Web Scraping
* Understanding of databases and API integration
How To Apply:
π© Interested candidates can share their CVs at umerrecruiter@funprimetechnology.com
Telegram Channelπ
t.me/jobsourceglobal
WhatsApp Channelπ
whatsapp.com/channel/0029VbAqbkw6hENvEJZLl22d
π€ Gemini 4 Carbon Tops AI Ranking
Artificial Analysis put it ahead of GPT-6 Astra and Claude Opus 5.5 in the overall ranking.
Carbon performs better on huge texts, math problems, and office tasks. Its biggest lead is in cybersecurity. It hallucinates facts half as often and cheats far less on tests.
Access is currently limited to cybersecurity professionals. Everyone else is promised access as soon as possible.
π t.me/techpsyche
Artificial Analysis put it ahead of GPT-6 Astra and Claude Opus 5.5 in the overall ranking.
Carbon performs better on huge texts, math problems, and office tasks. Its biggest lead is in cybersecurity. It hallucinates facts half as often and cheats far less on tests.
Access is currently limited to cybersecurity professionals. Everyone else is promised access as soon as possible.
π t.me/techpsyche
NIGERIAπ³π¬ Cubic Prints Africa is HiringGraphic Designer
π€Organization: Cubic Prints Africa
πLocation: Lagos, Nigeria
How To Apply:
Interested and qualified candidates should send their CV to: hr@cubicprints.africa using the job title as the subject of the mail.
Telegram Channelπ
t.me/jobsourceglobal
WhatsApp Channelπ
whatsapp.com/channel/0029VbAqbkw6hENvEJZLl22d
There is nothing as dangerous as becoming complacent under the assumption that you have time.
π§ AI mostly stopped inventing fake sources. Its flattery got stronger
Much of what we "know" about AI flaws is out of date. A Stanford team retested 50 well-known findings on 20 newer models: 44% no longer hold, and another 34% got much weaker.
In 2023, 55% of the sources ChatGPT cited in literature reviews didn't exist. Claude Sonnet 5.5 now fakes about 1%, GPT-6.1 Sol none. Some models still do: Qwen3.7-Flash invented 69%.
What lasted is behavior, not skill: agreeing with and flattering users, AI judges favoring their own answers, a left political lean, answering when they should hold back. Praise that follows what the user wants is now 1.85 times stronger than in the original paper.
The findings that expired lasted a median of about 15 months.
π t.me/techpsyche
Much of what we "know" about AI flaws is out of date. A Stanford team retested 50 well-known findings on 20 newer models: 44% no longer hold, and another 34% got much weaker.
In 2023, 55% of the sources ChatGPT cited in literature reviews didn't exist. Claude Sonnet 5.5 now fakes about 1%, GPT-6.1 Sol none. Some models still do: Qwen3.7-Flash invented 69%.
What lasted is behavior, not skill: agreeing with and flattering users, AI judges favoring their own answers, a left political lean, answering when they should hold back. Praise that follows what the user wants is now 1.85 times stronger than in the original paper.
The findings that expired lasted a median of about 15 months.
π t.me/techpsyche
Struggles and hardship are sometimes nature's way of keeping you away from the comfort zone.
Comfort zone is a beautiful place, but nothing ever grows there. It is a death zone in disguise.
Comfort zone is a beautiful place, but nothing ever grows there. It is a death zone in disguise.
Being offline is underated !
Messages, notifications and constant pings interrupt your focus and quietly disrupt your mood. Sometimes what looks like βstaying connectedβ often turns into a continuous stream of interruptions.
Sometimes the real productivity hack is simply stepping away from the noise and starting your day with DND mode turned on.
Happy Wednesday everyone
Messages, notifications and constant pings interrupt your focus and quietly disrupt your mood. Sometimes what looks like βstaying connectedβ often turns into a continuous stream of interruptions.
Sometimes the real productivity hack is simply stepping away from the noise and starting your day with DND mode turned on.
Happy Wednesday everyone
KENYAπ°πͺ Science Lab Limited is HiringJunior Software Engineer(2 Positions)
π€Organization: Science Lab Limited
πLocation: Nairobi, Kenya
Salary: KSH 80k
Qualifications:
* 0β2 years software development experience (professional, internship, or strong portfolio)
* Bachelor's in Computer Science, Software Engineering, or related field
* Proficient in at least one language (Python, JavaScript, Java, or C#)
* Basic experience with relational databases
* Familiarity with Git version control
* GitHub portfolio or academic project work
* Willingness to learn our stack
How To Apply:
forms.cloud.microsoft/pages/responsepage.aspx?id=1g8tH9N7CES9jDZhv1J17k6oBdps_y9HvNQGQ98zwqlUMVRXN1dEN0xMQUE4REhLOUNaWTJZWDk3Ui4u
Company Careers Page:
sciencelab.co.ke/pages/careers#dept-engineering
Telegram Channelπ
t.me/jobsourceglobal
WhatsApp Channelπ
whatsapp.com/channel/0029VbAqbkw6hENvEJZLl22d
Media is too big
VIEW IN TELEGRAM
π½ Mistral's AI built a live 3D map of the 19,000 satellites and bits of space junk circling Earth
Mistral says it gave its Mistral Large 4 model one job: build a real-time 3D tracker of the roughly 19,000 satellites and debris fragments in orbit.
According to the company, the model built a working app, tested it the way a user would, made it fast and passed strict verification. The app is called Orbital Trace.
π t.me/techpsyche
Mistral says it gave its Mistral Large 4 model one job: build a real-time 3D tracker of the roughly 19,000 satellites and debris fragments in orbit.
According to the company, the model built a working app, tested it the way a user would, made it fast and passed strict verification. The app is called Orbital Trace.
π t.me/techpsyche
β€1
You should master one skill before chasing others, because true progress comes from depth, not distraction. Once youβve built a strong foundation in one area, related skills become πππ¨πππ§ and more meaningful to learn.
SOUTH AFRICAπΏπ¦ Ergasia is HiringJunior IT Installer
π€Organization: Ergasia
πLocation: Freestate, Bloemfontein - South Africa
How To Apply:
webapp.placementpartner.com/wi/vacancy/?id=ergasia&vacancy_ref=Pta000487
Telegram Channelπ
t.me/jobsourceglobal
WhatsApp Channelπ
whatsapp.com/channel/0029VbAqbkw6hENvEJZLl22d
π§ GPT-6 Astra rewired a small AI from the inside and lifted its math score from 7.9% to 48.1%
MIT researchers handed OpenAI's GPT-6 Astra the raw numbers inside a small Qwen model, its weights, and asked it to make the model better at math. No training run, just edits.
Astra found the flaw that made the small model get stuck in reasoning loops, a problem Qwen's own model card warns about. It changed 0.016% of the weights. Instruction-following scores tripled too.
In other tests Astra looked at the raw weights of small game-playing networks and named the game each one was trained on. It also wrote small game-playing networks from scratch that beat ones trained for 100 million steps.
The team calls it weight-space programming in a blog post published today. Wharton professor Ethan Mollick: "Waiting for replication but this is a pretty big deal."
π t.me/techpsyche
MIT researchers handed OpenAI's GPT-6 Astra the raw numbers inside a small Qwen model, its weights, and asked it to make the model better at math. No training run, just edits.
Astra found the flaw that made the small model get stuck in reasoning loops, a problem Qwen's own model card warns about. It changed 0.016% of the weights. Instruction-following scores tripled too.
In other tests Astra looked at the raw weights of small game-playing networks and named the game each one was trained on. It also wrote small game-playing networks from scratch that beat ones trained for 100 million steps.
The team calls it weight-space programming in a blog post published today. Wharton professor Ethan Mollick: "Waiting for replication but this is a pretty big deal."
π t.me/techpsyche