Quando un’azienda riesce davvero a spiazzarci in positivo, vale la pena fermarsi e capire cosa è successo. Con Devstral 2 Small e Devstral 2 Large , Mistral ci è riuscita. Ha rilasciato due modelli specializzati nel coding in modalità agentica e, soprattutto, ha riportato al centro della scena qualcosa che sembrava superato: un modello denso da 123 miliardi di parametri . Un gesto quasi controcorrente, in un’epoca dominata dai MoE . Eppure funziona. 1. Perché Devstral è una release significativa Il punto è semplice: Devstral Large ottiene circa il 72,5% sui benchmark Verified per software development , superando DeepSeek 3.2 . In uno studio indipendente commissionato da Mistral, gli esperti hanno confrontato output di Devstral con: DeepSeek 3.2 → Devstral preferito nella maggior parte dei casi Claude Sonnet 4.5 → parità nel 25%, Sonnet preferito nel 50%, Devstral nel 25% Il risultato chiave? Devstral è competitivo con Sonnet 4.5 pur essendo molto più piccolo, m...
After four years of incredulity and not-so-gentle mocking, Bo Thide of the Swedish Institute of Space Physics and a team in Italy have finally proven that it’s possible to simultaneously transmit multiple radio channels over exactly the same wireless frequency. In theory, according to Thide, we could potentially transmit an “infinite number” of TV, radio, WiFi, and cellular channels at the same time over the same frequency, blasting apart our highly congested wireless spectrum.
Thide’s approach is rather simple. Basically, electromagnetic waves can have both spin angular and orbital angular momentum (OAM). If you picture the Earth-Sun system, spin momentum is the Earth rotating on its axis (producing the day-night cycle), and orbital momentum is the Earth rotating around the sun (producing the seasons). In standard wireless communications — radio, TV, WiFi — we only modulate the spin angular momentum of waves. For years, Thide had theorized that orbital angular momentum could also be added to wireless signals, effectively creating a spiral signal that looks like fusilli pasta; or, in the words of Thide, a “radio vortex.”
Now, in an experiment in Venice, Thide and his Italian colleagues have transmitted two signals at the same time, on the same frequency, over a distance of 442 meters (1450ft). Pictured on the right is the antenna that the team used. No, your eyes don’t deceive you: To create these radio vortices, all you have to do is make a cut in a standard parabolic reflector and twist it slightly. If you imagine a corkscrew of radio signals being continuously transmitted from the outside edge of the antenna, that’s effectively what’s occurring. On the receiving end, there are two “normal” TV antennae (Yagi-Uda) set apart by the same angle as the break in the transmitter. These antennae “decode” the vortex, and voila: Two radio signals transmitted over the same frequency.
It is hard to put into words just how significant Thide’s discovery could be. If the vortex preserves other aspects of wireless communications, such as multiplexing, then in the short term we could be looking at a wireless spectrum that can carry 10 or 20 times as much data. In the long term, as our understanding of orbital angular momentum grows, our wireless spectrum could effectively be infinite. To be honest, this is such a huge twist for wireless communications that the full repercussions are not yet known.
With radio and TV, and now cellular networks, wireless spectrum is one of humanity’s most valued resources. It is because airwaves are so clogged that companies like Verizon or Vodafone pay billions of dollars for just a few megahertz. If Thide’s discovery pans out, not only would wireless spectrum lose most of its value, but the trouble and strife surrounding LightSquared, international roaming, LTE rollout, white space wireless, and digital TV simply cease to be.